Slashdot Mirror
Apache vs IIS in Performance?
Greg Merideth asks: "In the midst of my proposal to universally sweep all of our company web servers together I was handed an advertisement from DELL by our CIO with a big circle around two machines that DELL sells with an interesting note. They are identical machines, literally identical machines but the claim is that the Linux 6.2 Apache based machine only supports 20k-100k hits per day while the Windows 2000 IIS based machine (remember, same config) supports 500k-1M hits per day. Now if DELL is claiming that NT, with the same config will out-perform Apache in serving web traffic, how am I supposed to convince my company that Apache and open source is a great way to go? They don't care about open source or Linux, just the performance that they will get from the machines. Where can I get -credible- data to prove that Apache can outperform IIS?"
You forgot one thing: you are an employee of Microsoft. :-)
Depends on if you mean CGI or ASP. CGI to CGI, Apache loses. But barely. CGI or JSP to ASP, Apache loses BIGTIME. ZDNet and several others have shown the same results.
Our site does use SSI's and on the low side each *.html request would mean a minimum of 10 "hits"--making 90,000 page views equivalent to 900,000 hits. So the first question is what are they calling a hit.
Also the box that runs this is little more than a slightly pumped up three year old desktop box (extra ram and faster disk)--I'm sure some /.ers have better game boxen than our server.
ACK
duh, heh, this is really easy to answer:
/index.html
a claim that apache can only handle 100k/requests per day? hmm.. ok, how about 100k requests in 110 _seconds_ on a loaded down file/web/internal gateway server with a p3-550?
1:00pm up 22 days, 4:12, 7 users, load average: 61.42, 25.27, 9.32
249 processes: 243 sleeping, 6 running, 0 zombie, 0 stopped
Server Software: Apache/1.3.12
Server Hostname: carbon.qspinc.com
Server Port: 80
Document Path:
Document Length: 2345 bytes
Concurrency Level: 1000
Time taken for tests: 110.564 seconds
Complete requests: 100000
Failed requests: 0
Total transferred: 261426140 bytes
HTML transferred: 234523450 bytes
Requests per second: 904.45
Transfer rate: 2364.48 kb/s received
Connnection Times (ms)
min avg max
Connect: 2 219 93004
Processing: 5 292 151
Total: 7 511 93155
yup, yup, time to sue for some flase advertising, eh?
Try it from RedHat patches/bug fixes. I think there's something that says that their 6.2 kernel performed badly under heavy load.. then they patched it... as far as I remember
With a load balancer working amongst 3 RedHat 6.1 , 2.2.15 kernel, Apache 1.3.6 machines, we server over 1.2 million page views a day, so 400k a piece. Dell never mentions a load balancer, but it is ESSENTIAL if you really want to run a serious business.
i'm AC because i really just don't give a shit any more. moderation is completely random and my karma can't go up anyways. my posts are all exactly what i feel at the time, be that pro or contra the current topic, and that seems to be an issue for the slashbots. fuck em. take any of a handful of flames/trolls/real posts from this article, and it's probably me. i'm stunned, if foolishly, at the amount of pure _bullshit_ posted about something as open, and for which there is a much freely available information, as apache.
apache 1 will probably not serve more pages per second on a tuned box than IIS. who cares? it will sure as fuck (see my response to post 1) push _way_ the hell more than 100k pages per day.
apache 2 is already much, much faster than apache 1, but, really, it just doesn't matter. it's all such ephemeral bullshit, arguing about what server pushes more bits down a pipe per second - in any real business, money and experience are the deciding factor. if your admins are NT guys, you're using NT, if not, you're using whatever they know. the real limiting factor of how well your site performs is how good your admins are and how much $$$ you have to spend on it - not one or the other, but both.
The question of which server to run may have already been answered before it was ever asked. I work in a shop that is primarily MS based development. We rely heavily on active server pages and VB scripts, therefore IIS was really the only way to go. My company has finally decided to break the chains of dependancy on MS and begin our web development in Java. Anyway, point is, find out if your web developers even CAN use Linux and Apache as a web solution, then compare where experience lies, licensing costs, etc and see where your longest list of PROs is.
http://slashdot.org/articles/00/07/05/0211257.shtm l
http://www.spec.org/osg/web99/results/res2000q2/
Yup, The apache processes run round robin (far as I know), and that means that you have to flush the CPU cache for each new connection (rotating the memory space).
Welcome to 1996! As part of this year's lineup of exciting new technological breakthroughs, we're bringing you:
Keepalive connections!
That's right! No longer does every connection spawn a new child, but we'll keep YOUR child open for as long as you're using it! How EXCITING can you get?!
buh.
and, to answer your question, 'not optimized for smp' is a misnomer because threads and forks are conventionally equivalent under most unices - the speed benefits gained by using a single, threaded web server under a well designed SMP unix kernel will be negligible compared to the child-process model.
.mp3. if that same monolithic mp3 encoder were threaded, it would work roughly n times faster under SMP, where n = num cpus.
:)
as a preforking server, apache is perfectly tuned for use on an SMP box. this is not true with a monolithic program such as a (theoretical) mp3 encoder which will not split a data file, use multiple processes to encode each chunk, and then reassemble them into a single
but, of course, you knew that.
the opposite is true (forking is difficult) under NT and, i assume VMSen.
If the hardware config is the same, you're cutting yourself short on the NT side. There will always be better performing, cheaper hardware that has NT support, but which hasn't 'scratched anybody's itch' enough that anybody has pleaded with Linus to incorporate it into the kernel tarball.
It's common knowledge that Linux hardware support lags by six months to a year. It's a fine OS for your old Pentium Pro machine, but don't waste the latest high end hardware on it.
I work for a company that is just breaking into the top 50 sites on the Internet, (Sorry I can't say which one) depending whose stats you go by. We run several racks full of Dells (2400 and 4400) Just about all are dual procs with a gig of ram. All run linux and apache. Then we use huge Sun Enterprise servers running the Oracle backend. Our biggest problems usually lie in keeping enough bandwith available and not the machines themselves. They can keep up just fine. Remember this though, you may start off with one or two servers but that will quickly grow. One or two servers for static, one or two for dynamic, one for graphics, MTA's etc. If you add all of those together licensing for MS becomes truly horrific. Linux is far easier to install and administer, it's reliable and cheap. The hardware requirements are considerably lower. Dell's support is okay but nothing to brag about. Their new 1U servers are slick. Keep in mind though that the cost vs. performance edge will always come out in Linux / Apache's favor. Especially on the new 2.4.x kernel.
That's total BS. IIS blows away Apache in serving dymanic content. I've done the comparos myself using ColdFusion Enterprise on IIS 5 vs. Apache 1.3.12. It especially holds the advantage on a multi-processor machine.
Why don't you ask Yahoo,Hotmail(yes hotmail) or CDROM.com what OS they run. By the way if all you are looking for is speed you should check out thttpd. Also,what is the speed of the link(now and future), the type of content being served. If you really want some speed and that is all that you care about check out this http://www.pdos.lcs.mit.edu/exo/ It is much faster than IIS on static content. newmaa2@rpi.edu
There are several web servers that may outperform apache on serving static content. Search freshmeat for possibilities.
Dell themesleves (with RedHat) have claimed record-setting performance using their "Tux" kernel-space web server daemon. (again for static content)
If you need to support more general content, then look at the flexibility in apache (and other Unix web servers) to support a wide array of different services and the ability to tune apache.
There's more to performance than just performance. Can you tune the server? How easily? Content management? Scalability to multiple servers/load balancing? Integration with database back-ends or java servelets? Security! Server management? Ease of upgrading? Paltform-dependancy? Licensing costs?
IIS ties you to one vendor's solution FOREVER. You can migrate an Apache site onto just about any O/S and hardware platform. This may or may not be important to your decision.
Depending on what you need and the skills your company possess, Apache may be an excellent choice or just OK.
I know you asked for benchmark data; Sorry I don't have a link to any and benchmarks are usually heavily cooked beforehand anyway (such as the Dell/RedHat example). IIS has a marketiing-savvy corporation behind it. Apache has the largest installed base on the planet and thousands of success stories.
There may be some info at the main apache web site, Apache Today, the apache week and oreilly sites (don't have url handy).
Oh, unless your firm is tied to buying only Dell, look at the offerings from VA Linux and the other linux vendors - I know VA can provide specific tuning suggestions.
there are 85400 seconds in a day. To say apache serves 20K - 100K hits is to say that it takes from about 4 seconds to slightly less than 1 seconds to service a hit on average. Doesn't sound right. They Dell should be more specific on how they measured this. Regards
But wouldn't the Linux 2.4 kernel HTTP Daemon change that?
He does know something you don't know.
"Don't trolls get tired?"
- How did you manage to exceed the PCI bus speed? The fastest that I've ever seen was a GSN nic running on an RS6000 with a 64 bit, 50 MHz PCI bus, transmitting at about 300 MB/sec, and that was 100% limited by the PCI bus.
I'm not being sarcastic when I ask this and I mean it to be a serious question: have any hardware companies (D-Link, 3Com, etc) considered creating an AGP NIC? Would that solve the PCI bandwidth problem?Alex Bischoff
---
Alex Bischoff
HTML/CSS coder for hire
Posted by polar_bear:
Um, just because China decides to utilize Linux, it doesn't mean that sweatshops in China are responsible for coding Linux - which is what I believe the earlier post was implying, that Dell utilizes Chinese labor to produce equipment.
China doesn't have to get permission to use Linux, and their use of Linux does not in any way imply support of Chinese politics or labor by other Linux users. That argument simply doesn't follow.
While putting a human rights clause on the GPL may sound like a wonderful idea, it really wouldn't be
very effective and would be counterproductive. Some might hope that by exposing people in countries with oppresive political situations to free software that maybe some of the ideology behind free software might rub off and cause change in those countries. (It's my hope that exposure to free software and the ideas behind it might spark some thought in countries like the USA as well....) That's probably wishful thinking, but everyone has to have a dream.
Posted by polar_bear:
This may be outside of the scope of the discussion, but one has to wonder what kind of marketing genius would put together a brochure like that. It's almost enough to make a person get conspiracy theories. Why would any retailer of any product advertise two identical products, with one touted as being lower performance? Do they display a great price differential?
This is very weird on Dell's part - they can't expect to make more money on the NT stuff. They can't expect to make friends in the Linux Community by dissing the Linux part of their server line. They can't expect to increase sales by claiming one is better than the other - so what did they hope to accomplish here?
Anecdotes are not as good as carefully created, unbiased tests. But since such tests don't exist, it's all there is to go on if you want to avoid the lies associated with marketting.
Don't label something "offtopic" unless you know the topic well enough to tell what's on topic.
Everyone is trying to quantify the problem as "well, for condition N, Apache does this but IIS does this". This is useless. It totally depends on the problem you're trying to solve.
I don't use anything near the default Apache config on really heavily hit web sites. I have it custom tailored to serve data for that site/application and to do it well and it might not work at all for a different problem/site.
Buy the solution that you are more comfortable working with. If you have a staff of largely UNIX oriented people, you will be able to bend Linux (or BSD, or whatever) and Apache to solve your problems easily.
If your staff is full of NT weenies, go with IIS. You'll probably be able to tune the system that much better and won't be delayed/sidetracked by people who have no experience with that system.
Raw and objective performance numbers for these 2 don't exist. Don't even try opting for one of them over another based on performance benchmarks. Go with the platform that your staff is more comfortable with, they will be the real cost, and they will be the ones who will have to deal with it. The price of the servers is negligable compared to that.
That should be 1000000/24/60/60 and that comes out to 11.57 hits a second.
Also you only did 100,000/60/24/24 so if that was the break down it'd be 28.93 hits per second.
Yay math!
The average throughput of traffic runs at around 120mbps, with peak times pushing 180mpbs. The last estimate of hits per day is around 25 million from around 200 thousand visitors. Hits are, of course, predominantly static (static webpage graphic art and the like) but a large contingent is dynamic, using PHP for most of the 'high impact' stuff and C or Perl based CGI for older goodies. Database goodies are handled on thier own machines.
Initial installation did cost a bit, as it was a conversion issue from sites running on single machines going to a unified environment in such a way that service wasn't interrupted. While static information can be served up easily, there was some rewriting involved for the dynamic pages (pushing DB lookups to remote machines, changing things so state data could be accessed from any of the machines in the farm, etc). Sites designed with distributed computing in mind would have none of those issues.
On the average, each machine serves up 1.3-1.5 million hits a day, and the webservers themselves barely break a sweat while running Apache running perhaps 200-250 processes at peak times (the only bottleneck worry is the TB data warehouse serving the machines right now). Thus, an affordable, fault tolerant, high performance arrangement using commodity parts that's reliable enough to let me go on vacation for two weeks or more without monitoring and not have me worrying if everything will be okay when I get back.
I'd suggest that you take any ads by Dell in the spirit they were written in... that of a Microsoft Lap Dog. The idea that even a default Apache install can only serve up 200k hits a day (which translates to a mere 3.5 hits a second!) is ludicrous and deserves scorn and ridicule.
--
rickf@transpect.SPAM-B-GONE.net (remove the SPAM-B-GONE bit)
"People will pay big bucks for the luxury of ignorance."
Maybe iPlanet is all repackaged IIS? ;)
--
rickf@transpect.SPAM-B-GONE.net (remove the SPAM-B-GONE bit)
"People will pay big bucks for the luxury of ignorance."
That said,if IIS is faster than Apache, which wouldn't surprise me all that much, it will not be -that- much faster, even serving basic static pages.
But answer me this, what is the pages served per minute from a machine that crashes at least once a day under any real load. Figure at least an hour to bring it back up an do post mortum.
Also, what does the inability of a sys admin to ssh into a box and do maintaince do to reliability?
That said, if you want out of hand performance for static pages, you can check out khttpd and such. I'm sure it would blow the doors off of static serving from windows. Course, in the end, static serving isn't all that challenging.
Chris DiBona
VA Linux Systems
--
Grant Chair, Linux Int.
Pres, SVLUG
Co-Editor, Open Sources
Open Source Program Manager, Google, Inc.
Amen. Life is too short to waste on idiots.
Reasons to NOT do Windows:
1. If they insist on a Windows solution they are probably idiots. Unless it's the only job available it's not worth it for that reason alone.
2. If they know so much, why are they hiring me? If they aren't going to trust my decisions at the start of a project, you can be damn sure they would be second guessing and micromanaging before it was over.
3. No craftsman should accept second rate tools.
4. If it's a team effort, working on a Windows project would mean having to bail out the clueless MSCE/VB kiddies they will stick on the project with you.
5. Corallary to #4. UNIX folks are more fun to be around. Windows people tend to be suits.
Democrat delenda est
Most i815E boards have this..
Example:
ASUS CUSL2
(32(bits) x 33,000,000(Hz)) / (8 x 1024 x 1024) = 125 MB/sec or so. In the real world, you rarely see over 95-100 MB/sec from a single device.
Admittedly I was using Task Manager to do this, but IMHO I _should_ be able to kill the process (after all, the only process that should be protected against kill are the shell and the idle task, and perhaps a few more, but NOT IIS).
Also, what is the point of having an administrator's account if he can't do anything?
In those cases you use your friendly "kill" command (in the resource kit or pskill from sysinternals.com) and either kill inetinfo.exe,
.. but) "Administrator can basically do whatever the hell he wants"
Well explain then why (logged in as Administrator, not as a user in the admin group) I have on more than one occasion had to reboot a server because I couldn't kill inetinfo.exe? It even gave me a nice little warning box stating that I didn't have proper permissions to kill it. Well what good is an administrator's account anyway if he can't kill whatever process he needs to kill? I even checked Users and Passwords and there it was, in plain english (forgot the exact phrase
All it takes is a little tuning. Check this out:
click here
I don't think that my old graph is updated anymore, but those machines were/are only 200MHz AMD's on old motherboards with 256M of RAM each with a NetApp F630 for back end storage. The spikes are webcrawlers, mostly Google nowdays. A little calculation will show that if we could have attracted "peak hour hits" all day, it'd be 3.3M hits between the 3 machines, so 1.1M per machine...and those machines hardly break a sweat! Those machines used to regularly handle around 30 hits/sec each. I recently spec'd out a system for a company that had a $20k price tag, is completely Linux and could do about 10 times that load. Course, we're talking mostly static stuff. If you factor CGI's into the equasion, ModPerl is far faster/better than ASP or whatever NT will have you use, so that raises the edge Linux/Apache/Perl has even further. Tell your Boss or whoever is questioning you that statistics are wrong and that you know best or he wouldn't have hired you. Apache/Linux is the way to go period. Ask him to justify 3 admins to keep up with NT's retarded problems and random reboots vs buying extra hardware and keeping 1 admin. Extra staff (esp. admins) is expensive, hardware is cheap.
"E Pluribus Unix"
Well, I beg to differ. I've had to reboot Win2k after changing ips. What I have said about Win2k is my personal experience with it compared with what the Win zealouts espouse as positive changes for it, which are not entirely true.
I have never had to reboot linux after changing ip's.
Why do I want to run 7 copies of Q3A? I dont even want to run one copy.
Dell - who cares what they think. I got the hardware, installed what I want on it.
And btw, a wonderful degeneration of things by suggesting that I'm pulling this out of my ass. No, I haven't.
-- There is no sig line, only Zuul.
> With only a few tweaks, most (if not all) of
> the security holes in IIS are covered
Reboot after all those patches too, then be sure to reapply all the service packs you might have to.
> As for Apache, you always have to upgrade the
> entire server (can you say downtime?)
Yes, that would take a lot of downtime...
killall httpd ; $pathto/httpd
or
/etc/rc.d/rc3.d/S85httpd restart (for the RedHat people - maybe others)
lots of down time there, how many milliseconds was that?
All the configuration for an upgrade of anything on linux can be done while the existing server still runs. You can even test it by running it on some other port in the meantime. Then when ready, kill the oldversion, start the new version. Downtime of seconds. How many minutes will it take to reboot an NT machine for changing something
like the ip address?
> and maybe even the Linux/BSD kernel (can you say
> MORE downtime?).
Again, how long will it take to reboot the machine for a kernel upgrade? 2 minutes? 3 minutes? The same amount of time it would take to reboot a windows machine for any number of upgrades/changes/etc. But the linux/BDS reboot is only necessary to upgrade the kernel, and not to upgrade the software running on the machine.
AS far as rating downtime of NT vs Linux for maintenance reasons, I think your reasoning is all wet.
yacko
-- There is no sig line, only Zuul.
> Ah, but you forget something: most, if not all,
:)
:)
> of the IIS holes are already documented.
I highly doubt that all the holes in IIS are documented. We just haven't found them all yet. I'd bet that we continue to see more bugs from both IIS and Apache, that's the inherent nature of things.
> There is ONE service pack for Win2K.
So, for any and all Win2k bugs, there's only one fix? How long did holes stay open then while people waited for that SP? And how many holes are currently open waiting for the next SP? To count the patches in this manner is apples and bushels. Apples for linux cause each patch is counted, where as with windows, the SP is simply a bushel of oranges - all different size oranges - and you dont know how many oranges are there. It's not an equal measure.
(and you STILL have to reboot after a service pack!
> Also, regarding changes to NT, haven't you seen
> 2000 yet?
Yes, I still have to reboot it if I'm changing seemingly simple tcp/ip settings. Change from DHCP to static? reboot. Change from one workgroup to the next? reboot. lather, rinse, reboot.
We've recently received some new puters in the office. All Dells. They all had windows on em. There's 9 of em in the webbie room and 12 in the programming room. All the ones in the programming room had windows removed and Linux installed. Even with that difference in numbers (9 win2k to 12 linux), I'm willing to bet that the webbies have rebooted more than the programmers. (hmm.. would be an interesting thing to try and track
Yes, Win2k doesn't crash as much, according to the webbies, so that's good. I take that as a good sign. And of course, workstations vs. servers would be a little unfair.
GRanted, this conversation we're having has little to do with IIS vs. apache, but more so the *nix vs. Windows NT (core NT - inc. 2000) stability.
-- There is no sig line, only Zuul.
Offhand it sounds almost like the SCSI bus is being reset. Have you checked your kernel messages? Obviously, check the termination and everything as well, but you've probably done that already. Could just be flaky drivers...
Look at the web benchmark more closely, it shows Linux outperforming Freebsd not the other way around. Freebsd is outperforming Linux in compile times and nfs however.
-- You can be a geeklord too
It's been thought of, but typically, the types of machines that need this kind of serious throughput don't have AGP slots. When the idea came up at my company, it was shelved because we had lots of PCI experience and no AGP knowledge so we'd have to start from the bottom of the learning curve.
An SGI Origin 2000, with multiple XIO busses, is the usual solution, I think. Some of the government labs have clusters of O2ks connected by a GSN network.
Actually, we are talking about megabytes per second. My situation involved a single GSN nic (6400 megabits/second) and his involved 8 GigE cards (8 * 1000 megabits/second).
I'm still not sure how he got to 370 MB/sec. 4 32bit/33MHz PCI busses maybe (AFAIK bigger PCI busses aren't available on x86 machines, but I'd love to be wrong), or maybe the GigE cards had some kind of "diagnostic mode" transmit generated by the GigE hardware that doesn't involve PCI bus activity.
Or maybe you're right and the original poster's talking about megabits while I'm talking megabytes. Oh well.
For static content it's all thttpd.
I did not ask Dell, but I wonder if you did.
Muchas Gracias, Señor Edward Snowden !
Now let's see: a day has 86400 seconds. Let's limit that to daytime hours only, say 50k seconds. 100k hits in that time makes 2 hits per second, on average. If Dell claims that's the limit of their hardware, well, my Atari 800XL would be faster as a Web server.
OK, I know that a hit and a hit don't necessarily mean the same thing... but speak to your boss and let him just apply some common sense here.
Union Pacific Railroad is in the midst of converting everything over from NT to Apache running on Solaris. The NT webservers were way too unstable, crashing several times per week.
Where I work, Microsoft salespeople came into our group, and started telling us that IIS is behind 70% of ALL e-commerce sites. We laughed them out of the building. We asked to be explained why to use their product, we did not ask them to lie to us about made-up things. iPlanet had just told us a few weeks ago that THEY have 70% of the eCommerce market. Hmmm... That adds up to 140%, someone must have been doing math in the vicinity of a black hole again...
Don't believe the marketspeak - it's usually fabricated, unverifiable bullshit. Tell them "Show me the money - set up identical Apache and IIS servers here and prove your distorted figures."
I adminned IIS servers for 5 years, and just in the last year started doing Apache. At first, the lack of a GUI is intimidating, but once you get familiar with the configuration file, you find that Apache can do all kinds of things that IIS never dreamed of...
Andrew
http://www.acooke.org
Umm....in regards to OpenBSD, you should also show your boss this...
..but..psssstt..."
From skyper@SEGFAULT.NET Thu Oct 5 11:28:01 2000
Date: Thu, 5 Oct 2000 18:40:16 +0000
From: skyper
To: BUGTRAQ@SECURITYFOCUS.COM
Subject: obsd_fun.c
"hello hello obsd team. my obsd box panics every few seconds.
what the hell is wrong?"
"oh ? really ? hmm...out of space in kmem_map ?"
"YES. you know about this bug ?"
"yes. some kiddo is running a DoS against your box.
we fixed it in 2.7. the kernel runs out of memory if you
flood it with arp-request.
"oh. ok. i wont tell anyone. ive heard of some
other exploits for obsd. maybe its time to change
the 'Three years without a remote hole in the default install!'
to '3 years without releasing an advisory' ?"
(Exploit snipped due to ethics)
I don't know why both Apache and IIS benchmarks are so slow... My 100Mhz 486 benchmarks 2.2M hits per day....
--
If you want a really fast (and simple!) web server for Unix, try Jef Poskanzer's thttpd. thttpd is entirely select based, single thread, memory mapped IO. Really fast. mini_httpd is also great if you want a zero-configuration web server, but it's not as fast. Jef's a brilliant old school hacker.
A common configuration on big web sites is to use Apache to serve the complex stuff, and thttpd to serve static images.
I sent in a few links to LinuxToday for a story about the SpecWEB99 numbers from Dell recently. Here is this link to the resulting story.
The bottom line is that Dell has been submitting systems to the Special Performance Evaluation Corp (SPEC) to be included in the SpecWEB99 results. Dell using the Red Hat TUX 1.0 web server has kicked the butt of all other submissions with the single exception of the IBM eServer pSeries 680 with 12 CPU's running Zeus. The Dell / Red Hat Tux 1.0 (8 CPU) numbers come in at 6387 simultaneous connections. The IIS/5.0 systems highest score is on an IBM Netfinity 6000R (4CPU) comes dragging in at 1582.
Hope this helps.
Tony
My God! It's full of Voids!
This is not a new claim by far. You'll need to do a little bit of research on how their benchmarks were made. Specifically, IIS _is_ very good at serving static content, since it pretty much caches everything in memory, whereas Apache does not.
To simulate the same, partner Apache with the newer kernel-based khttpd, which is a kernel-space HTTP server, which will pretty much blow everything away for static content. Configuration is simple. You basically set it up so that static content (ie. flat HTML files, images, etc...) are served off the "accelerated" khttpd server, and everything else is sent to the Apache server.
Benchmarks are like voodoo--bad voodoo at that. Dell is showing you their voodoo from a given viewing angle, but anyone can grab it, rotate, flip, mangle, to their hearts content and get about whatever results they'd like to see.
In order to avoid this, spend an hour and figure out what _YOU_ consider to be important, and then test based on your values, an identical hardware configuration, using both IIS and Apache. These are the only results that should matter to you.
Don't allow yourself to be spoonfed what someone else thinks is important. Remember, they're a vendor...it's their job to make money off of you, not necessarily sell you the best, or right solutions.
-db
I don't know if it is NT4 based. That was one of the errors on a partner site but I've run into similar ones very often in my online shopping treks. We run Linux/FreeBSD/OpenBSD here so I don't have to deal with some of that stuff. Of course you also see sites that don't anticipate MySQL/whateverSQL going down and they've got the same errors. But these Windows based servers throwing errors are on high profile sites. Rather annoying as a shopper too.
Error Occurred While Processing Request
Error Diagnostic Information
Error attempting to get the client (Client ID = '1078348').
A problem was encountered trying to access the system registry. Error number 1450 occurred.
The error occurred while processing an element with a general identifier of (CFAPPLICATION), occupying document position (11:1) to (15:48).
Date/Time: 10/06/00 11:29:56
Browser: Mozilla/4.0 (compatible; MSIE 5.01; Windows NT 5.0)
Remote Address: 000.000.000.000 (to hide meself)
Template: F:\www_customers\techbargains\index.cfm
Please inform the site administrator that this error has occurred (be sure to include the contents of this page in your message to the administrator). -------------------------------------------------- ------------------------------
---------------------------
Error Report
MBuser.user GetUserInfo3 Failed: /Default.asp- --------------------------
----
Please hit your BACK button, jot down the exact page, operation you were
trying to perform, page item and link that produced this error and include
it in the e-mail to us of the problem, along with the system diagnostic
messages listed above (copy and paste the text above into your message).
E-mail this information to so we can remedy the
situation for you quickly!
-------------------------------------------------- --------------------------
----
Thank you for your response!
--------------------------
EVENT DATE: 9/18/00 - 11:35:57 PM
ERROR NUMBER: -2147467259
ERROR DESCRIPTION: Timeout expired
ERROR SOURCE: Microsoft OLE DB Provider for SQL
Server->MBUser.User.GetUserInfo3@MBWEB1[version 1.0.0]
HTTP_HOST: www.removed_host.com
PAGE:
QUERY_STRING:
SESSION CRUMBS:
MBPI=009769EC88A911D493AC00508B8FA300; MBST=t;
MBPA=0;
MBSKC=n; MBSV=0; MBSA=0;
MBSI=009769EC88A911D493AC00508B8FA300
USER AGENT: Mozilla/4.0 (compatible; MSIE 5.0;
Windows 98; DigExt)
-------------------------------------------------
Yeah, I was thinking the same thing. A Cobalt Cube can server up to 10 million objects a day, meaning files, e-mail, and web pages. So, if it was web only, that little puppy with just 16MB of RAM and Kernal 2.0.x will outperform the Dells. And it is cute too.
About 50% of which are dynamic PHP or Perl pages, 40% of which involve hitting MySQL, which is running on the same server, or an external IMAP server. You see, this machine is mostly an IMP webmail gateway.
And every hit is 128-bit SSL.
unet.brandeis.edu is a 2-year-old valinux box with a 350MHz Pentium II and 128MB RAM.
Load average is usually around 1.5.
Uptime since January 3rd, when management allowed us to turn the servers back on after a pointless y2k shutdown, is 100.0%.
We have new hardware, but haven't taken the time to cut it over because the old stuff is working.
An identical machine is serving as our web proxy. For one day's statistics, see http://brandeis.edu/~rcgraves/squidc.html
I'm not making this up.
That's a 350MHz Pentium II with 128MB RAM serving 700,000 http hits, 5.3GB data, with Squid 2.3. Load average is typically around 0.5.
has done more than 100k hits a day, while cpu average was around 2%...
Then again, here's a company selling a kid's version of Linux
preinstalled. I certainly wouldn't trust them to do my Apache config for me.
The only real way to tell is to buy both and test them exactly as delivered.
It's called Level 2 cache and is on processors.
--
Delphis
My personal experience has not shown that IIS is faster than apache. I would say that apache has 1.5x the performance of IIS.
Dom
an astute observation, but mayhap you're assuming too much objectivity in his case. is it not possible that he really doesn't care which is better and just wants data to prove that apache r0x0rs or something? :)
lies, damn lies, etc.
-dk
-dk
Dream with the feathers of angels stuffed beneath your head.
So, Dell is claiming that an Apache server running on their hardware is only capable of serving up about one page per second, max?
"Great men are not always wise: neither do the aged understand judgement." Job 32:9
i tested out a box that was a pii-450 with 128 megs of ram running linux. i got it to pull about 3500 static pages/second. this fully saturated the 100Mbps network i was on. the load on the machine was about 40, but it was keeping up with the requests. the program i was using is called ptester. it may not have been the most accurate results, but it did show allow me to tweak apache's config to give the best results. doing some math, that gives me a theoretical max of 302,400,000 hits/day (60*60*24*3500). i don't know any site that would pull that much. dynamic content would pull that down quite a bit.
20000hits/day is like one hit per 4 seconds. i know many sites that pull tons more than that.
Why read the article when I can just make up a snap judgement?
You are spot-on with the rest of your comment, but the "Oh by the way I'm a Unix guy" *is* a valid argument. Unless there is an overwhelming need otherwise, go with what you know. Learning something new is great and all, but do it on your own time, not by installing something into production and saying "Cool! Now, everybody just hang out and take a smoke break while I reinstall because I screwed it up the first time."
:-)
For instance, I am in a mixed NT/*nix shop (mostly Solaris and AIX) and the company is interested in dropping linux in as DNS or Samba servers, but haven't given the go ahead because of lack of Linux experience in the admin dept. Note that they *want* to do it, and the boss is satisfied that linux is stable/secure/whatever enough, he just doesn't want to be woken up in the middle of the night while someone on the other end says: "well, I *thought* I knew how Linux handles slices, but, umm, we need to restore from backup".
Btw, I post all this in the assumption that you wrote the "member of the elite crowd" line facetiously (sp?), cuz as an "elite unix guy" I am here to tell you NT is easy to use but hard to master, esp. if you are trying to start with no windows experience at all. I'm not saying it's better or worse than Unix, but it just *feels* weird. And this is after I installed cygnus tools and vi
I can't say as I'm familliar with the process of gaining access to Solaris's source code, but I assume (and you even make this point in your post) that you can do the same with NT. I remember reading a post from a gentleman who goes to school at a uni where they have all 5 cd's full of NT's source code. He even complimented Microsoft on it's readability. So I don't think that part of the argument holds true. I was just refering to the fact that it is completely different from unix in how it is put together and where things are. Call me a command-line commando, but I long for my three Eterms with bash, not cygnus-bash with ported vi. It doesn't feel right, and the file system layout is all wrong. Again, I'm not saying one way is bad and the other good, just that I'm very, very, used to /etc, vim, and HOW-TO's when it comes to setting something up, not technet, regedt32, and a prayer ;-) If I missed your main point, please feel free to correct me and make me see the error of my ways...
Your point being? My point is that even with dynamic content, the average pentium III server today could server at least 50 req/sec (counting very low). This is 4.3 million requests in a day. If you get 20 million requests in a day during 11 hrs (US east/west coast market hours) that is 500 req/sec. Now, that might be hard for one server if the data is dynamic, but definitely not a problem for static data. Also the numbers quoted were 20k to 100k and 500k to 1M. That is NOTHING, trust me.
Using the example 11hr, 1 million reqs is just 25/sec (on average). To be able to serve 20k requests in a day, you need (hold your hats) to be able to pump out a page in 4.3 seconds! WOW, that's fast!
The scale quoted in the Dell article (if that is in fact what it said) would have been a problem in the distant past with CERN running from inetd on a Sun 3/80. Maybe.
It's easy easy for any PIII box (if you have the bandwidth) to pump out millions and millions of (static) requests per day.
20k? Ha! I have never in my life seen a box that was that slow. Even CERN httpd on a 496 could do many times that.
There's a big difference between saying that something is easy and saying that it's accurate.
I'd never want to do that job. It doesn't match my skill profile at all.
I also consider myself foolish whenever I allow myself to be convinced by a sales pitch. Occasionally it does prove to be accurate, but the frequency is such that depending upon this is folly. I don't even believe OUR "salesmen", and they aren't even officially trying to sell me anything. It just gets engrained in them.
Caution: Now approaching the (technological) singularity.
I think we've pushed this "anyone can grow up to be president" thing too far.
It might help if you called DELL themselves and asked how they arrived at their ratio of 10:1 performance differential. Maybe they based this on the results of impartial tests performed by some third-party group with no axe to grind.
Like MindCraft.....
Your position would be more credible if you also boycotted MicroSoft for being based in a country which has been repressing people in mezo and South America for at least 100 years, and most vigorously in the last 25.
I wonder if you'd stop to consider why a human rights activist would concentrate on a distant country's excesses while ignoring those of a closer, putatively more democratic country?
Could be that drawing attention to Tibet takes it away from Guatemala, El Salvador, Chile, Brazil, and the support western governments (and in particular, the US) lend their bloody dictator/clients?
Just a thought.
"There are four boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order." Ed Howdershelt
First off, you should recognize that the Dell server hardware is optimized for Win NT, not for linux. I have several running both, and I can tell you from personal experience that while Dell's linux guys are nice, dedicated OSS people, they are NOT the guys who get to decide what the hardware will be. They simply try (and lately, they've been succeeding) to back-port Red Hat to the machines Dell is already making.
Second, as far as I know the IIS advantage is entirely based on serving static pages. You need to know what you're serving, and if it's static pages exclusively those numbers might be meaningful... but probably not, as a good caching service like SkyCache (whatever they are calling themselves now) or Akamai can give you better performance to the rest of the world than on-site IIS machines can ever achieve.
Third, even if somebody has done a test using some technique other than static pages, such a test will not reflect the performance you're going to get. Your php and perl, your cgis were not developed to test httpds, they were developed to facilitate your business goals. Your mileage WILL vary.
So, to get conclusive results: Buy two identical cheesy "white box" PC clones. Get fast Pentium !!!s, yes, but assembled at the local Pakistani computer mart, not at Dell or Gateway. Now load IIS and Apache. Time how long it takes to get the machines running. Record number of crashes AND number of reboots from the time you get the machines in the door to the time the test has completed. Load your actual pages and cgis etc. Get a load tester (anyone care to recommend one? I use a room full of humans, which may be cost-prohibitive for you) and TEST WITH YOUR REAL LIVE DATA.
Now, take the results of these tests (including the "how many crashes" records, and how long it took to get the systems running correctly) and write up a little report. Point out the endless parade of security hazards involving IIS. Point out that Dell servers are not the highest-performing linux boxen. Point out the flexibility of Apache in accomodating new languages and techniques. Point out that IIS is losing ground to Apache on the net (must be SOME reason, right?).
Then bottom-line it with cost. Leave out hardware costs entirely (they will be nearly identical) and compare free to Micro$oft.
Works for me.
--Charlie
That's not quite true, IIS does have a kernel-mode
driver, probably to implement an optimized
disk->network path. Given that Linux 2.4 has the
same thing I don't see how Linux people could
complain.
and apache. besides, check the math:
500,000 hits * 30KB = 15,000,000 KB = 15 GB
20,000 hits * 30KB = 600,000 = 600 MB
you'll probably get at least 3 servers, so that means you'll be able to handle 18,000MB with apache and linux.
Why are Windows mags testing Linux in the first place?
(Slightly off topic but it dose get to the hart of what the Troll only glazed.. every Linux vs NT benchmark seems to come under attack on Slashdot)
Windows mags are read by Windows admin. They don't subscribe to such mags to be told about OTHER operating systems.
The only need information on ONE operating system.. Windows.
They don't need to know if Windows is better than Linux only how to make Windows NT run at it's best.
If they did prove Linux was better than Windows NT then what?
Microsoft faithful would unsubscribe in discust. Linux converts wouldn't need the publication anymore.
Why? They do it to keep readers.. becouse if they switch to Linux.. they won't need the Microsoft publication anymore...
I don't actually exist.
...China has chosen Linux...
That's right! They chose it. It's out there, free for *anyone* to choose it.
It's not like Linus or Bob Young went off to Beijing, briefcase in hand, trying to rustle up some business for the sake of the American dollar.
Also, Linux and other GPL software should NEVER have a 'human rights' clause in it. How fucking hypocritical would that be? Mr. Stallman probably cringes at the thought (at least he should) that his ideal of FREE(dom) software would be free except for those who we don't think worthy of it.
Human rights issues aside, if you don't want a segment of people to choose and use your software, don't GPL it.
Blech. Signatures.
Yeah, but you know what?
That clause is directly related to the software which the license is for! Imagine that!
You can't use the GPL code in a non-GPL'd program. The GPL doesn't say you can't use the code if you paint your house green, pick daisies by the side of the road at four in the morning, the author doesn't like your hair, jerk off in front of the girl next door OR kill, abuse, and generally shit on the population of a nation you're in charge of. Why?
Because it's irrelevant. A driver's license doesn't say you can't watch reruns of 'BJ and the Bear' because it's irrelevant. A hunting license doesn't say you can't eat pie on Sunday because it's irrelevant. The GPL doesn't say you can't engage in human rights violations because its IRRELEVANT. LIKE THIS THREAD.
Blech. Signatures.
Look up freedom in the dictionary. What does it say?
I thought so.
Simply put: If you place restrictions on who may use 'Free Software' based upon something as arbitrary and subjective as 'Human Rights' then it ceases to become 'FREE Software' and becomes 'Free-only-for-those-we-want-it-to-be-free-for Software'
Like I said, you want to release *your* software under the MrEPL or whatever that says 'If you beat dogs you can't use this software' then fine. Do so.
You are completely missing the whole point of the GPL, which is FREE for EVERYONE software, regardless if we like them or not.
As for most Tibetans, I doubt if most of them really give two shits to the wind about software of any kind. Again, not the point.
Also, by choosing to use Linux and other OS/FREE software, they have accepted into their society a speck of freedom that would otherwise be controlled by some corporate and probably American entity, like Dell or Microsoft (to make a lame attempt to get back on topic. ) Which sounds better to you?
Go back to hugging trees and saving baby seals or something.
Blech. Signatures.
It's Apache vs. IIS, not Linux vs. NT/2000. Apache runs on many Unix variants, and even Wintel, but it's most tested and stable on Unix by far. It would be interesting to see a BSD/Apache, Linux/Apache, Win2k/IIS comparison, and any other good combinations anyone would like to come up with. Yeah...Linux can be beat in plenty of areas....but in general, serving is a strong point. If it were losing that badly in some "test" of serving, I'd be skeptical of that test as well...and I am not even a hardcore Linux fan. (I'm still sticking with my BSD :)
Let the stat's (and methods) speak for themselves. Any true scientist would never ask "How can I prove that Apache is faster than IIS?" We need to ask "If I compared similar installs of Apache and IIS under the same conditions, how do they perform?" Anything else would be unreasonable.
Let's look at Dell's testing methods. The same hardware is meaningless if they applied tweaks to NT but left the Linux installation to defaults.
> As for Apache, there's always
d b/middle.html%3Fvendor%3DApache%2520Group %26title%3DApache%26version%3Dany
d b/middle.html%3Fvendor%3DMicrosoft%2 6title%3DIIS%26version%3Dany
:-)
> bugs/holes/orifices being found, and Apache has
> a new patch at least twice every month. To my
> knowledge, IIS5 has been patched ONCE (in Win2k)
> SP1).
Hmm, well let's see what bugtraq has to say...
Number of security holes listed for apache: 10
http://www.securityfocus.com/frames/?content=/v
Number for IIS: 59
http://www.securityfocus.com/frames/?content=/v
(Links done like this because the posting engine kept inserting spaces in the URL)
Note that some vulnerabilites for both are actually caused by 3rd party add-ons, not Apache or IIS. But, this sure tells me alot about which one I would trust.
> One more funny thing: All the manhole covers in
> Boston say "BSD". Kinda reminds me of BSD's
> place in the computer world.
Yeah, BSD only runs the highest volume web server in the world, the highest volume FTP server
in the world, and Microsft even borrowed from their TCP/IP stack for Win2K. But yeah, you're right, it sucks.
-Wintermute
Didn't NASDAQ just get hax0red through IIS?
Yup.
And not only did they hacked through IIS, the hacker himself admits that the exploit was not an existing, patchable exploit, but one he discovered himself.
--
Seriously though... why the hell would you buy
a dell for a high end server?
Any sane admin konws that REAL hardware
is built to order out of GENERIC parts instead
of some hacked up, castrated crap that these
big corps call computers.
Build to order with linux in mind and you'll get
higher performance than if you put MS on the same
machine.
Build a machine without linux in mind and you'll
get what you researched.. a crapshoot.
Go figure.
Friends don't let friends buy Compaq's. (Dell/Gateway... same same) You want a good computer? Build it yourself.
This box is serving about 300K hits per hour! So the figure of a max of 100K per day is far too low!
The DELL box should easily handle millions of hits a day. The data as shown on your sheet cannot be correct; not for IIS and neither for Apache.
ms
D'oh!
Try approximately 1.16 hits per second instead . . .
They're taking their dog to get its two shots before it's too late. You're taking your dog there too, right?
Apache.org probably gets more than 20k-100k hits per day. Slashdot CERTAINLY does.
I've built 2 sites running Apache which eclipse 1M hits per day (one uses Zope behind it for the dynamic layer (with PCGI). That copes with 1.5M per day without coughing.
I'm very skeptical of the claim and would love to see some data.
Secondly, apache sites don't show ANY source code when you add "+.htr" to the URL.
Take your pick but let your manager make it HIS job on the line if its the wrong way to go!
lolol in other words which is better optimized for porn ?
beauty is only a light switch away
I recently built a server using a Serverworks 370DL3 motherboard, and a Mylex ExtremeRAID 1100 DAC960 card. On the driver page for the card, it reports an issue with the DAC960 firmware that can cause the card to stop responding under high load. Could DELL be using rebranded MYLEX cards? It might be worth a try to see if the MYLEX firmware would work on the DELL card. Check out http://www.dandelion.com/Linux/DAC960.html for more info.
If you can't figure out my address, just drop me an e-mail and I will explain.
5 years ago, NCSA on a 486 and FreeBSD 1.1.5 did 117K hits/*hour*; Dell should get better drugs.
rofl
:)
Me as a non native speaker always was keen on knowing the meaning "unbiased".
Thanks
This is not true.
IIS runs in a user mode process 'inetinfo.exe' and does not take down the OS when it dies. ASP pages can also be executed 'out of process' inside a Microsoft Transaction Server package that runs as another process 'mtx.exe'. This way, the application will restart itself and not even bother the core process.
I don't know where you heard that it's in kernel mode, but it's very much not so. I've heard rumours that this might happen at some stage though I've seen nothing concrete.
I used to be one of these people that didn't want to admit that NT is sometimes the right tool for the job. However, based on your comment, I suspect that you don't want to admit that Linux is sometimes the right tool for the job. I could be wrong, so feel free to flame me on that point :-)
Have you ever upgraded an Apache installation? I've never had to recompile the kernel to upgrade Apache. I've never had to upgrade the entire server to upgrade Apache. Where did you get this information? First-hand experience? If so, why have you had to upgrade the entire server or kernel to upgrade Apache?
When I do an upgrade of Apache, I test out the new version on a different port. This means that I run both versions concurrently - I don't want the customer to have any downtime.
When I'm done testing, I bring down the old Apache server and bring up the new one on the right ports - all on one command line. This means that the customer has less than one second of downtime during an Apache upgrade.
Is this possible with IIS? Can you run two concurrent versions on different ports? If not, how do you limit downtime and do proper testing when upgrading IIS?
>>what is the point of having an administrator's account if he can't do anything?
It's Microsoft.
Right. IIS is better at serving static pages from a 4-cpu 4-nic RAID-0 box.
For anything reasonable, Apache/Linux tends to be faster. Considering the inherent robustness of Apache (pre-forked processes), Apache is really the winner even if it were a bit slower.
Thanks for the links. Surprising, not that Apache was much better than IIS (about 4:1), but that these large professional sites are down about as much as /. ( with new undebugged perl scripts yet ;)
I disagree. Your test is maybe the only statistically valid means of measuring customer satisfaction. If IIS somehow is responsible for a disproportionate number of bad sites, then there is a problem with IIS.
Nope. ;)
On a casual hacker's day, 60% of the sites he hacked were on IIS and 30% were on Apache.
The interesting ratio is 60/20 to 30/60, which implies that IIS is 6 times as likely to be hacked as Apache. ( as well as 4 times as likely to be down
>>but don't waste the latest high end hardware on it.
You mean like the new high-end IBM mainframes???
Sorry, but NT only runs on cheap hardware.
>>or somebody doesn't know what they are doing...
;)
But, how do you find out what is going on.
>> or what hardware to buy
or what software to buy
Grader's grade: D-.
"that I could have" is excessively wordy in an online context.
"the" for "they" is a minor typo that does not interfere with the legibility of the post.
"modifier out in la-la land" diagrams like Dell supports. Looks like effective use of the language.
Anecdotal evidence for the "technical superiority" of FreeBSD. Looks like you run into a few things where the years of stressing and banging on BSD matter. In this kind of area, benchmarks are worse than useless, since they never measure the pile-on, pile-on effects. I suspect that you have some effects where everybody gangs up on the server at its weakest moment, like what hapens when you put ten tons into a half-ton pickup. Between BSD and Linux, expect Linux to have a better "best-case" ratio and for BSD to have a better "worst-case" ratio. The best tests will always be real-life loads. Good Luck.
There is a good reason for Apache's stability, and it is at the expense of speed. The primary httpd spawns a bunch of secondary httpd's which actually service the requests. Configured normally, these child processes serve some number of requests, and then self-destruct. Works wonders for buggy modules and memory leaks.
It's not necessarily being wrapped up in linux superiority, (though it does come off that way).
It's been my experience that sales material (which is what his boss gave him) is horrendously inaccurate.
And I'm not just saying "one system may be tweeked while the otehr isn't", I'm talking *huge* discrepencies between marketing literature and reality. Sometimes numbers will be off by orders of magnitude.
He's right not to trust marketing literature, and to seek independent statistics. In fact, I'd say it's part of his job. It's clear from his question that he is biased. So is Dell. And for all the faults of the Slashdot crowd (they are biased as well), I am fairly certain that if W2K/IIS really is an order of magnitude faster than Linux/Apache, as Dell says, that fact will be evident, either in the form of posts giving results that say it's true, or lack of posts that say it isn't.
Not that it matters with such a small number of hits, but you cannot just divide by 24*3600; it's about the peaks. Most hits are during office-times and usually there's a peak early in the morning (depending on the type of website).
0x or or snor perron?!
well apache is not the fastest web server, the apache group doesnt pretend to be. But i know for a fact it can handle over 1 Million hits per day. I've managed many sites, from some getting 100 visitors a day to ones getting 5 million a day. Apache has never failed me, it's actually been more ofa limitation of the OS ive had to deal with tuning-wise. As for OS , ive used Solaris (x86 and sparc) and various BSD's. What you need to look at is not only the web server but the OS for any performance eqaution as well as netowrk and hardware. This type of thing is too hard to boild down, but those claims i know are false.
this space for rent
Of course IIS will do more hits.
Hits as in "number of times you hit the box in frustration."
http://www .de ll.com/us/en/hied/topics/products_iis5_pwrap_paw_w in.htm
For Apache
http://www .de ll.com/us/en/hied/topics/products_iis7_pwrap_paw_l in.htm
It's totaly messed up. If looking at the linux page, A 600mhz and 128meg could do 100k-500k hits a day while a 650mhz with twice the ram can only do 20k to 100k!! Someone screwed something up along the way. The header on the questionable table looks like a typo. When comparing the IIS and Apache pages, you'll notice Dell's rating is about the same for the same basic config hardware.
MarNuke
-----------
"You can't shake the Devil's hand and say you're only kidding."
It should be noted that most web sites use alot of static data. Especialy the images. For every web page pulled, there are probably 3-4 images pulled too. Tux should speed up most any website, although some by not much.
Say you have 10 connections to an apache server. That means there will be at least 10 apache processes running. It's up to the OS to distribute CPU time among those processes.
Like any other program, Apache does not have to know or care about SMP at all. It just has to divide up work that can be done concurrently into threads or subprocesses.
Hands in my pocket
Netcraft says that www.olympics.com was running Domino-Go webserver, as you might expect.
... how many times does a heavy-weight OpenGL screen saver have to be selected on a server before someone will think about the load? :-)
GUI: the power to screw your server faster and more efficiently than before.
...your own performance tests. The only way to make sure that one OS platform will out-perform another OS platform with your specific web-applications, in your specific environment, is to do the tests yourself. Usually large vendors, such as Dell, will be happy to ship you a couple of eval boxes of the same configuration for testing purposes for a couple of weeks. _Usually_ all they request you do is sign a form saying that if you break 'em you'll buy 'em, and that you'll purcahse the box at the end of the evaluation period. This isn't a big deal if you know it's the hardware you want, but need the time to choose the OS to run on it.
Also, of course, make sure you take into account other features besides performance. Security, scalability, and reliability (just to name a few) should all play a part in OS platform evaluation.
In short, run what works best in your environment.
ALG
I can't speak for Linux, but I've got 10 Windows 2000 Advanced Server/IIS 5.0 boxes which each process over 1.25 million requests per day without breaking a sweat. The only thing that's close to being utilized it the disk sub-system. They are beefed up Compaq ProLiant 1600's.
ALG
Doesnt slashdot run apache? How many hits a day does slashdot average?
i saw the same thing in a dell ad this morning. here are the specs for one i think he's talking about. it's a PowerApp.web 100 1U machine: pIII-750, 256MB ECC SDRAM, dual 10/100 Intel nics, 2 9G ultra160 SCSI HD. the only difference is Win2000/IIS vs. Red Hat 6.2/Apache. they claim 500K-1M hits for the Win2000, 20K-100K for linux.
Slashdot runs apache and I bet it gets 500k to 1M hits a day just from trolls refreshing to get first post ;)
As one of the Senior Sysadmins who has worked on sites like MSN-Linkexchange and Hotmail, sites that serve *ENORMOUS* amounts of traffic with FreeBSD and Apache. A *single* PII-450 with 128MB of ram could serve well over a million hits per day. The trick is concurrency, and well, 64MB of ram would probably suck, but thats still ~50 active connections at a time under FreeBSD. If you're getting 50 hits per second, and your request takes a second or less, hey, thats a lotta hits :)
There was a recent /. article about Tux, it's a http kernel addon for linux .. it is supposed to speed linux' webserving up a LOT...
Ask Ingo Molnar About TUX
that url should get you started :-) i think specweb results are more credible than what some sales monkey dreams up...
That Dell brochure is talking from its hat - Apache on a *Palm Pilot* can manage 20k hits/day down the serial cable. Linus oughta sue :-)
:-)
If you look at the SpecWeb99 results to date, the number 2 and 3 slots seem to be held by Red Hat / Linux / Tux running on Dell kit, 8 and 4 CPU boxen respectively; the one system to beat them was fairly serious $100k iron - a 12-CPU IBM RS/6000 server - running the infamous Zeus. So maybe Tux is the best webserver?
The real joke is the fact that it is dyanmics that matter; any piece of junk can serve a ton of static content all day, you only have to haul it off the disk and push it down the damn wire.
At work, we front end on the cheapest Intel-based rackmounts we can find, using Apache 1.3.12 on FreeBSD (more secure than Linux) - since 90% of what we serve by volume (99.9% by server load) is dynamic content, the load average on those boxes barely flickers up to 0.05 on a busy day.
IIRC Walnut Creek used to run ftp.cdrom.com on a single 4-way Xeon box (Apache / ProFTPd / FreeBSD), and push about a terabyte a day of downloads. No worries.
Whatever you set up, load is unlikely to be an issue.
Yeah, propose this to the boss.
It'll give him an excellent opportunity to can your ass, then buy an even larger IIS server with the money you wanted to piss away by playing test rigger for four weeks.
And this would be the Sound Business Decision.
"Consider yourself a member of a virtual corporation with Mr. Torvalds as your Chief Executive Officer." - Linux Advocac
There is NO SUCH THING as Linux 6.2.
"The price of freedom is eternal vigilance." - Thomas Jefferson
On one single-processor 600MHz Pentium III box I happily serve about 70 hits/sec of mostly static content, or about 250,000 hits per hour during peak.
The main bottlenecks I've had are I/O related; the hits are spread over about 4.5 million indivdual files. But a half-gig of RAM takes care of that pretty well. I expect that the box will top out at about 110 hits/sec.
How is TUX a "bleeding edge webserver"? As far as I knew, it was just the kernel HTTPD on steroids, which will always be an ugly hack on a monolithic kernel.
Or is it not KHTTPD in disguise?
--------
Life is a race condition: your success or failure depends on whether you get the work done on time.
Because my bog standard old Dell 1300 machine is happily churning out 300k PAGES per day at dslreports.com, and taken at peak hour speed, can do 1m PAGES per day no problems at all.. Pages are produced in 20ms to 200ms during this, and they are all dynamic.
Meanwhile, in microsoft MSDN performance document base, they consider an ASP page to be too slow only if it takes 5000ms or more to generate..
There are lies, damn lies and statistics. Then there people actually using these things to get work done.
OMG...
Your siting linux for downtime. This is truely hilarious...
A kernel update requires a machine reboot... (5 minutes on our compaq servers which takes entirely too long for the raid controllers to initialize). An apache upgrade? Compile in different directory... shutdown old apache... start new apache... remove old apache... cp over new apache to old... ON THE SAME COMMAND LINE. (You don't even have to adjust your startup/shutdown scripts).
NT... change ip address (reboot)... apply hotfix for latest NT bug discovered 5 months ago (reboot)... IIS dies for no reason...won't come back up (reboot)... sneeze wrong way (reboot and dsk check)... 60 days pass (reboot).
Yes, I am exagerating a little bit, but not by much...
"You should always go to other people's funerals; otherwise, they won't come to yours." -- Yogi Berra
Unfortunately, I do not believe there is a clear cut answer on which is better. Many people have commented that apache serves dynamic content much better then IIS. Licensing issues are also a concern... clustering and smp machines.
However, we know from the mindcraft tests that the Linux kernel's lack of a threaded TCP stack has hurt it in the comparisons.
Following kernel threads you might note the IO subsystem overhall that has taken place. You would also note the threaded tcp stack implementation on its way. Apache itself has undergone changes since these "tests."
Enhancements are coming to level the field in the areas Linux/Apache are weak in... it might be a matter of simply waiting and then reviewing a set of truely independant benchmarks and analysis.
The answer is very complicated... you simply have to look at your needs and weigh both sides.
"You should always go to other people's funerals; otherwise, they won't come to yours." -- Yogi Berra
NT4 is not dead...
It is finally stable...
It took a number of patches, blood, sweat, and tears... but it finally has uptime.
You want me to Windows Two-Grand now? Fraid not, maybe in another year.... I have issues with using bleeding edge technology on production servers.
So for now... *sigh* (reboot)
"You should always go to other people's funerals; otherwise, they won't come to yours." -- Yogi Berra
For comparison, my PowerBook can do about 70/sec running WebSTAR on MacOS 9.
In other words, ANY server on the planet can handle that level of traffic -- for static pages. Beyond static pages, it has nothing to do with the server software, and everything to do with 'the other stuff' (DB, middleware, etc.)
-Mark Kriegsman
Clearway Technologies
Actually, a $0 software investment represents exactly zero percent of the long term cost of the server. I'm sure they explained to you how to calculate a percentage of something somewhere in school.
AC also said
You mean they would rather hire college dropout NT zealots, such as yourself, instead? How come the only zealots we read about here on Slashdot are Linux zealots? Probably because the ones doing the writing about Linux zealots are MS zealots with an axe to grind trying to make all Linux users all look like MS-bashing freaks.
How juvenile.
If the $25,000 described as the expense of the software is 1% of the TCO for the system, then the system must have a TCO of $2,500,000. If the lifetime of the system (the only major variable that I don't have here) is 2 years (as would be implied by setting the total cost of software at 2 years because you'll have to pay $25,000 every couple of years if you buy a Microsoft solution) then it's highly unlikely that the bulk of the TCO is administration. At $200,000/year (the cost of a full-time employee with overhead and whatnot that we used at (big company name deleted)) that means that you could pay 6 full-time staff to just to administer that system.
That's too many people. A good, stable system would require two part-time administrators (a primary and a backup, typically, and there won't be all that much administration to do if everything is working properly) so you've still got to account for at least $1,700,000 of your estimated total cost of ownership. I know it eats electricity and A/C and backup tapes, but nearly two megabucks worth over two years?
So, your figures don't add up unless there are other conditions required. I've seen safety-critical systems cost that much, but I've not seen anything that suggests that anything you are talking about is anything other than a heavily-used Web server. Care to add some details so we can determine what the TCO of the competing solutions really is?
Oh, and anyone that thinks a company is going to ignore a $25,000 expenditure simply because it's small WRT the total cost of a project is living in some kind of fantasy world.
My digital watch could handle 100,000 hits per day. I tried ab (ApacheBenchmark, comes with Apache) on my workstation, which is significantly more powerful than my watch. ;-) The config:
- --------------
- --------------
- --------------
Athlon 850/512MB PC133/Adaptec 2940 (U2W LVD)/9GB Cheetahs. This was going through localhost since Apache, and not the network, is in question. Running RH62, default kernel, and more importantly, default Apache installation, with a slight system load (desktop crap).
Bottom line: 112 to 128 seconds to fulfill 100k requests. Page was the Piranha (static) login page. Taking 120 seconds as an average, that's 72,000,000 hits per day under constant load, on relatively cheap desktop hardware.
Using -k (keep alive) with 128 tasks, this even better: 71 seconds, 1396 hits/second, or 121.6 million hits per day.
-----
primus:~# ab -n 100000 -c 128 -k localhost:8000/
This is ApacheBench, Version 1.3c <$Revision: 1.38 $> apache-1.3
Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Copyright (c) 1998-1999 The Apache Group, http://www.apache.org/
Server Software: Apache/1.3.12
Server Hostname: localhost
Server Port: 8000
Document Path: /
Document Length: 1314 bytes
Concurrency Level: 128
Time taken for tests: 71.620 seconds
Complete requests: 100000
Failed requests: 0
Keep-Alive requests: 99078
Total transferred: 164082524 bytes
HTML transferred: 131419710 bytes
Requests per second: 1396.26
Transfer rate: 2291.02 kb/s received
Connnection Times (ms)
min avg max
Connect: 0 0 112
Processing: 1 91 6992
Total: 1 91 7104
-----------------------------------------------
primus:~# ab -n 100000 -c 32 localhost:8000/
This is ApacheBench, Version 1.3c <$Revision: 1.38 $> apache-1.3
Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Copyright (c) 1998-1999 The Apache Group, http://www.apache.org/
Server Software: Apache/1.3.12
Server Hostname: localhost
Server Port: 8000
Document Path: /
Document Length: 1314 bytes
Concurrency Level: 32
Time taken for tests: 126.877 seconds
Complete requests: 100000
Failed requests: 0
Total transferred: 160411228 bytes
HTML transferred: 131409198 bytes
Requests per second: 788.16
Transfer rate: 1264.31 kb/s received
Connnection Times (ms)
min avg max
Connect: 0 6 194
Processing: 12 33 740
Total: 12 39 934
-----------------------------------------------
primus:~# ab -n 100000 -c 128 localhost:8000/
This is ApacheBench, Version 1.3c <$Revision: 1.38 $> apache-1.3
Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Copyright (c) 1998-1999 The Apache Group, http://www.apache.org/
Server Software: Apache/1.3.12
Server Hostname: localhost
Server Port: 8000
Document Path: /
Document Length: 1314 bytes
Concurrency Level: 128
Time taken for tests: 112.513 seconds
Complete requests: 100000
Failed requests: 0
Total transferred: 160476992 bytes
HTML transferred: 131463072 bytes
Requests per second: 888.79
Transfer rate: 1426.30 kb/s received
Connnection Times (ms)
min avg max
Connect: 0 27 166
Processing: 21 115 716
Total: 21 142 882
-----------------------------------------------
primus:~# ab -n 100000 -c 512 localhost:8000/
This is ApacheBench, Version 1.3c <$Revision: 1.38 $> apache-1.3
Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Copyright (c) 1998-1999 The Apache Group, http://www.apache.org/
Server Software: Apache/1.3.12
Server Hostname: localhost
Server Port: 8000
Document Path: /
Document Length: 1314 bytes
Concurrency Level: 512
Time taken for tests: 124.992 seconds
Complete requests: 100000
Failed requests: 0
Total transferred: 160443308 bytes
HTML transferred: 131435478 bytes
Requests per second: 800.05
Transfer rate: 1283.63 kb/s received
Connnection Times (ms)
min avg max
Connect: 0 115 93103
Processing: 4 304 268
Total: 4 419 93371
--- Bigger bits, softer blocks, tighter ASCII.
I have worked with NT and Windows 2000 since NT started and I have been doing linux since 93 and I can tell you that a good distro with Apache is just as fast as Windows 2000 doing static pages. Apache and PHP beat the crap out of anything ASP can do. Also Windows IIS does not use SMP. If you do dynamic pages it might use SMP but IIS does not. Also IIS can only have 250 concurrent connections. If you have modem users hanging off your IIS server the 251 person gets dropped. Apache will go to 1000. So neither IIS or Apache use SMP. IIS only does well if you are testing on a local lan and you are testing small static files that fit in the cache. So for most people IIS or Apache will do the job. Do you want to pay the Microsoft Tax or not.
By Linux 6.2 Apache based do you really mean RedHat 6.2 running Apache? What kernel? What version of Apache?
I suggest you first start out by phoning up DELL and asking them to explain their methodology for the tests that they are basing the claims on, and the versions of all the software used, and the tweaks applied to all the software. I'm pretty sure Win2k IIS will beat Linux 1.2.36 running Apache 1.1.x as well. :)
There are also several other variables that aren't mentioned. Static pages? Dynamically generated? Via what language?
After talking with DELL about their methods of testing, then ask them to lend you a demo unit for a couple of days to run your own tests on. "Sounds interesting. I would like to pursue your solution further for our server needs, but I would like to have a server to test some of our applications on prior to committing to the purchase." I haven't found a vendor yet that won't ship out a test unit.
Then, when you have the test unit, install Win2k/IIS and test the results, then install Linux/Apache, and test the results, and go by those benchmarks.
And just out of curiosity, what type of hardware was it anyways? :)
"(remember, same config)"
The hardware config might be the same but
software config can't be...
AdFuel
Pr0n industry uses Apache due to budget concerns. (Not that I know much about it otherwise....)
Have your managers read this headline, and see how it turns out...
Not really related to performance... Well, maybe the world's economic performance *cough* *cough*.
Don't sweat the petty things. But do pet the sweaty things.
Greg,
1 hit per second is simply not possible, unless there is a serious problem with the apache configuration. Apache running on the most minimum hardware possible should easily do better than that.
If your boss understands logic at all. Once you prove that the test is flawed, no value can be given to any of the results. NEVER trust benchmarks, especially if they come from someone trying to sell you something.
Benchmarks == BullShit
Which also shows that, on a casual hacker's day, 12% (60% of 20%) of the sites he hacked were on IIS and 30% (30% of 60%) were on Apache.
NEVER use statistics. It does no good to anyone.
Try to remember that Apache doesn't have a PR department whose salary depends on pretty glossy documentation. Apache just has to work. That's it. That's why it's ubiquitous.
Yeah, there's is a lot of OSS and Apache propaganda. Try to sift through the bullshit successfully. It's not a lot of effort to get the boxen you need and set up Apache for a firm comparison.
I've used IIS, Apache, and NES (may they be struck down), and Apache worked best for my scenarios because I needed the systems to be working all the time. IIS enjoyed a few random lockups every 5-6 days for me. I can't tell you how *happy* I was trying to troubleshoot the problems.
One argument you can use is that it's free to try Apache, you can always go out and buy IIS.
Jayson's $.01
Never go to sea with two chronometers; take one or three.
In case anyone wants to know and doesn't have the Dell Brochure, here's more information:
Dell October 2000 Catalog Page 31
Dell PowerAPP Server Appliances
Comparisons (These are the ones that confused me the most):
Intel PIII 700
9 GB 7200 RPM Ultra3 SCSI HDD
Windows 2000/IIS has 128 MB Ram
Red Hat 6.2/Apache has 64 MB Ram
The Windows 2000 Machine is rated for 20K-500K hits per day
The Red Hat 6.2 Machine is rated for 20K-100K hits per day
Next Comparison (Above)
Intel PIII 750
Two 9GB 7200 RPM Ultra3 SCSI HDDs
Windows 2000/IIS has 256 MB Ram
Red Hat 6.2/Apache has 256 MB Ram
Windows 2000 Machine rated for 500K-1M hits per day
Red Hat 6.2 Machine rated for 10K-100K hits per day
Odd, I don't remeber talking to you? Hmm... If it makes you feel any better I don't like the people here either... Drop me a line, I'm curious if I have spoken to you...
FYI- Adaptec finally released a binary for that damn RAID controller. At least I (and everyone else) can recompile my kernel now.
Hrast
Dell has some serious problems if all they can manage is 100,000 hits per day with IIS. Running that very same program on my 500MHz laptop I can pull that many hits in less than a minute.
> The server might be easy enough to administer that an IT flunky can keep it going
There's a stack of dead simple apache interfaces which are much eaier to use than the IIS4/5 MMC plugin.
Most notably commanche: www.commanche.org - there's a stack of others, but commanche is the best. IMHO anyway.
Yould could also use e-smith [www.e-smith.net] for a dead simple web administered [no Linux or Unix knowledge required] Linux based machine.
An ISP or NT shop would prefer the Cobalt boxes, which have SSL Apache, PHP, and ASP support. Cobalt boxes are also configured through a web-based interface.
Both of the above solutions are far more simpler than NT IIS4 / W2K / IIS5. Tell your box you've had it with archaic, non-reliable technology that costs you money to fix, and money to train people to use. Go with e-smith or Cobalt.
> For the price of an MS-IIS license
You mean an NT/2K license, client access license, and internet connector license. But we get your point.
Also consider the likely move to non-perpetual licensing in the near future.
> Would you cut Microsoft the same slack if they came out with an advisory saying "in order to make a Windows machine secure, block all Internet services" as you advocate for Linux boxes?
They do. Read the Technet security guidelines for running IIS in a `secure' fashion - the document suggests disabling nearly all other services.
BRW: Inverted commas for `secure' because nothing is 100% secure, not because I'm a Linux / Windows troll who uses terms like Windoze / Open Sores.
> Tux is a kernel module that integrates the html protocol directly into the kernel.
:-).
*HTTP* protocol directly into the kernel. But we know what you mean
> Many changes which required a restart in NT4 now only require a restart of the specific process, just as in Linux/BSD.
I think our definitions of `many' are a little different. So you can change your IP address without a reboot. Yay. Try modifying anything* else in control panel / network.
Can anybody speak for apache's performance on Win2K?
I suppose I'm not too threatening, presently, but wait till I start Nautilus
Well I don't have any data to back this up but IMHO, Win2k will be faster on a Multi Proc X86 type machine for spitting up static pages. It threads better. Period.
Now before y'all moderate me down for saying good things about the evil ones, there a few things you will want to bring to your bosses attention.
1. Are you only serving static pages?
2. Kernel 2.4 will probably improve this considerably
3. You would proably be better off using a cluster of cheaper 1U(~$3000) machines and a load balancer(running linux)
4. IIS crashes if you look at it funny
5. If you're going to spend 50 grand or so anyways, check that IIS server versus an Ultra Enterprise 250r. No contest.
You could also ask your boss about just what he expects Microsoft Partners(tm) to say in their adds. I would have been shocked if they claimed linux to be better at anything, probably violate's the contract.
Aaron
Where did you get the idea that IIS runs in kernel space? It is a service, granted, but it is _not_ a device driver, nor is it part of the main OS kernel.
Maybe you know something that I don't, but I'm pretty sure it runs in userland.
Yes, since parts of IIS are (AFAIK) handled in the NT kernel.
For a true comparison, compare IIS with the Linux kernel-internal webserver (also only for static pages). I bet Linux would win then.
Capt. Ron
crazy dynamite monkey
Which is a PIII 550 with 256 MB RAM and I just checked the logs and we average around 400K Hit's a day and we serve approx. 300K Files / day. And the box load is very low and it runs other services as well.
So I am a bit surprised how they came to that number.
Michael
If you want to e-mail me, use my PGP Key.
> From the original question:
> They don't care about open source or Linux, just the performance that they will get from the machines
> They're only concerned about performance. The license fee shouldn't be a deciding factor unless the two systems are otherwise quite similar.
Indeed.
However, as others have pointed out, it should be made clear that there are many ways to measure 'performance'. Dell, and the poster's management, seem to imply that 'hits per day' is the way to measure performance. Another, is up time.
I wonder what the figures for 'hits per year' would be, taking into account that when a machine is down it has zero hits I'm sure there are many different metrics for performance which you could use to counter the 'hits per day' metric. Don't know if it'll be enough though.
Max.
[1] I don't suppose it'd make enough of a difference; but you can figure into it the additional down time required due to additional maintenance effort that might be required, and weight the down time in the calculations.
Max.
And if you're lucky, one of them might even know enough to install NT and IIS and get the site up and running.
This is probably redundant by now, but...
If one team of people configured both machines, then one of them probably wasn't set up right. We should stop doing web benchmarks against computer hardware and start doing benchmarks against money.
Two teams should be set up representing 'experts' in either IIS or Apache. Each team is given the same amount of money to create and configure the best web machine they can with that money. The money will be charged for hardware, software, and hours of labor (for setup and configuration). Then those machines could be benchmarked (with static and dynamic content) against each other.
This would allow proper configuration of each machine and the benchmark would be against the real bottom line --Money.
I know I'm risking a troll rating here, but shouldn't the question be "Where can I get -credible- data to prove which is better?" If you don't have access to the information that Apache kills IIS, how do you know that it is, in fact, better? I'm not saying you should take Dell's advertising at face value, or that IIS is better, but you are presupposing the answer to the question that you (admittedly) don't have the answer to. I suspect Apache would be better on many, if not all fronts, but I don't have any data to back it up either.
Agreed. As Sherlock Holmes says to Watson: "Twist theories to suit facts, don't twist fact's to suit theories." (or something like that).
"I'll take the red pill. No! Blue! AAAaaaahhhhhhhhh"
- Monty Python meets the Matrix
Microsoft sets up their web server to run closer to the Windows kernel. Many slashdot users note, less accusationally, that Microsoft has made technical decisions that hurt more users than they help. But not many suggest that Apache add an option to behave similarly -- with huge warning labels around it, of course. PHB's who (often incorrectly) put speed as a higher priority than security and stability choose NT/IIS because it's faster.
--
The shareholder is always right.
...the parent post to this is full of it.
I'm probably feeding a troll here so moderate accordingly.
Using your wonderfully thought out logic, would you mind explaining why Linux 2.2.x and apache will outrun W2k and IIS5 on similar hardware configurations when serving up entirely dynamic content? Hmm? I don't believe there is any 32bit texture mapped 3d eye candy being generated there either...
(Hint: Think it might have something to do with the way static HTTP requests are handled in the kernel in W2k and not the 2.2.x linux kernels?)
If you honestly make decisions in your job based on the kind of reasoning exhibited in your post then I would have to call into question your fitness to be an admin.
I'm not a Linux fanatic either. I use several versions of unix as well as NT/W2K daily in my job.
Karma? Who cares...
You're right that we need more TPC-W data. It's coming. The important point is that it will be a lot more meaningful for real-world web servers than SPECWeb99.
SPECWeb99 is a terrible benchmark for making a decision of which web server you want to run for a real business. Almost all of the pages in the SPEC benchmark are statically generated. A more useful benchmark is TPC-W, which essentially is Amazon.com (see the TPC website for more info). TPC-C is the standard benchmark for measuring databases, and TPC-W is likely to become the standard for web servers.
How exactly does one test a nic? just curious as to what apps you'd need to do this.
Erian
-
-
This Post has been brought to you by the letter "E".
Maybe they need to make better servers? I've got a lowly PPro 200 here with ancient SCSI hardware that currently pushes right about 1M hits a day. Workload avg's .2. Granted, it runs FreeBSD, but Apache on FreeBSD and Apache on Linux are not that different. Throw NT and IIS on that system, and it would probably slow to a crawl.
Just what is it that Dell is smoking?
There are certain pieces of hardware that perform better under NT and there are pieces that perform better under linux. For instance, a 64 processor Alpha will have better performance under NT because the SMP has been heavily developed in the kernel, where linux is currently lacking in a solid SMP core. Compare that to a 486 with 4 MB of RAM, Apache will run like a dream and NT won't even boot. Then you need to factor in the RAID controller and make sure that linux can support it as well as NT blah blah blah. Personally I laugh at companies who shell out thousands of dollars to buy software that can do the same thing as free open source software can. In some cases, NT makes sense. In most, Apache comes out the clear winner. No software is perfect for every configuration, and it's possible that NT makes a better performance server using this hardware config. Just remind them, though, that the extra "performance" if Dell is telling the truth will cost their pocketbook quite a bit more in software.
"You'll die up there son, just like I did!" - Abe Simpson
There is no reasonable defense against an idiot with an agenda
:wq
The problem with buying the computer with NT pre-installed then dumping it later on, is that you also are buying the service/support contract with Dell. I doubt that they would provide much support for someone who buys an NT computer then installs linux later on.
For the normal consumer it probably isn't a big deal, but for a company looking for a complete solution, it's a huge issue.
Come on, guys, you guys can't recognize a troll this obvious? I'm very disappointed in you!
"Prejudice is wrong; you should hate everyone the same."
Point well taken. If you split each process off, then are you not running similar to Apache and take a performance hit there also?
:)
I completely agree that the process can be killed, (albiet by adding a Unix command), however, just this morning I had an IIS box that was completely brain-dead, and did have to be rebooted. However, it was a developer problem (one that should not have been on a production server!) and had more to do with ODBC and transaction server than IIS, but that was the train of thought I was in when writing the response.
I would still take a performance hit in place of stability and reliability.
Definitely interesting, loadable modules do rule, I was not aware that Apache was moving to a multithreaded architecture, and thank you for an informative response.
Err... 20K-100K hits per *DAY*? That's just over 1 a second tops... Apache on Linux can handle 50 times that without breaking a sweat...
B1ood
Note to self: pasty-skinned programmers ought not stand in the Mojave desert for multiple hours. -- John Carmack
Good post. You've managed to actually apply some thought to the issue. Nice to read, rather than the knee jerk, linux rulz/MS suckz
At this level the OS isn't as much of an issue. And as everyone seems to be pointing out (with a Linux slant of course) there are more factors than just speed to be considered. What is the network design? Are these boxes just sitting out on the net or is there a load balancing box or boxes? What kind of network are they on? Are they multi-homed? What kind of peering do you have? Are you in a datacenter? What kind of rack costs are you dealing with? Static or dynamic? Big database backing these boys up or no database, or tiny database? What database are you using? Do you have a SAN on the back? Are you a portal? Will you be providing services like e-mail, chatrooms, etc?
From the question it seems like you've already made up your mind without truly considering the alternatives. What steps have you already gone through to verify that a Linux/Apache based system is indeed what you want? Why are you not into MS? I've been in the GlobalCenter (Borregas) facility and pound for pound (well Sun e450's and those huge, raid array, dual power supply, big iron, database systems weigh quite a bit) VA Linux seems to be in the lead for dark side rack mounted nifty looking boxes.
Good luck
That may be true when the numbers are more similar. But, just using these numbers, for the number of Apache based boxes you'd need to buy to get the equivalent of NT performance you could buy an extra box for fault resilience and a bigIP. At least.
Are you going to use these boxes darkside?
I just want to give props to IO Completion Ports on Win32. I know everybody thinks that MS is a bunch of jackasses but IO Completion Ports make this all possible (speed not jackasses). They're freakin k-rad.
>That should be 1000000/24/60/60 and that comes >out to 11.57 hits a second.
The math is correct here but you divided 10^6 instead of 10^5.
>Also you only did 100,000/60/24/24 so if that was >the break down it'd be 28.93 hits per second.
100000/60/24/24 = 2.893 as he claimed.
Matyas
"A society can be judged by the ways it entartains itself."
How exactly does quake performance translate to better web service performance? Please give me a technical answer backed with logic.
$var = <STDIN>
$var =~ s/\\$//;
$var = <STDIN>
$var =~ s/\\$//;
this is slashchomp
Ok, here is the error in your logic: Quake pushes a machine to it's limits by giving your graphic card as much work as it can take. The only way Quake performance is even remotely similar to web serving is multi-player quake, but it is like apples to oranges. Another factor you have to look at is what an operating system is built to do. Linux was not built to play quake fast, neither was windows, but more development time has made windows a better platform for quake, including graphic driver development. Some companies won't work with the open source movement to build great linux drivers, and some release linux drivers as an afterthought.
The real factors that determine good web service are: disk i/o performance, memory management, network configuration, bandwidth, security & many more things. But graphics cards don't factor into this.
Don't get mad that people here are backing Linux, and don't get the idea that everyone here is a zealot. This whole article is about which web service package is better, and if Linux comes up lacking, the Apache development team will probably work on it, and catch up. But the question about what is better is being asked now.
$var = <STDIN>
$var =~ s/\\$//;
$var = <STDIN>
$var =~ s/\\$//;
this is slashchomp
Even if your arithmetic were correct (it isn't) your assumptions aren't. Web traffic doesn't arrive in a linear distribution over time.
On the other hand, my experience is that Linux/Apache is many times more stable than IIS/NT, takes a lot less time and money to develop for, and comes with much better support. I'd fire a CIO that picked IIS/NT without a very good reason.
"Oh oh we dont have this feature yet but it WILL be in next version....."
Umm, that's called vaporware, and it is NOT unique to Linux.
Consider this: You can go out and download the latest development version of the Linux kernel to test the newest features for yourself. Try doing that with a Microsoft product.
"Try that with xyz" I never said a thing about Microsoft or how it relates to the topic.
Fair enough, I guess I sort of overreacted on that one.
The main thing I want to say is just that it doesnt seem fair or rather helpful to this guy that a kernel that isnt even production level does what he needs it to.
The point I wanted to make was that, though development software may not suit your needs, Open Source provides at least some measure of defense against vaporware.
If Apache promises threading in version 2.0, you can generally verify they are at least working toward this goal. With non-Open Source projects, or those with closed development cycles (like IIS), you can't. You are at the mercy of Marketroid claims.
This is not to mention that, often, it can be more costly to migrate platforms (say from *nix to NT or vice versa) than to simply wait for a feature you need to be supported by your O/S.
I don't know about you, but I think I'd be a lot more comfortable waiting if I had some assurance the features I needed would actually be there in the next release.
http://w8.hitbox.com/wc/index.cgi?C1 640 7540
The first reaction when hearing ANY statistic should ALWAYS be to ask :"What was wrong with the data/tests?"
Now of course, thats easy to forget when they come out in your favor...very easy. I will admit, I have done it myself.
Its a question of "what does the data really mean?". Its very easy to accidently design and run a test that ends up testing something other than what you wanted to test.
-Steve
"I opened my eyes, and everything went dark again"
> Now... Apache on NT vs Apache on Linux - that
> would be interesting. Even then, there would be
> the questions of performance tuning - which
I haven't looked at the web page for a bit for Apache/W32, but I definitively remember the Apache team stating that there was a "lot of progress that needed to be made" on the Win32 port of Apache.
I'd wager that means a comparison of Apache on NT vs Apache on Linux isn't even justified; not to mention the lengthier development time of Apache on *nix vs. NT (including optimization, etc.) simply because Apache on *nix is more mature and robust.
-- Talonius
My reality check bounced.
...that a majority of the posts here don't directly address his question? Most posts have instead fallen back to the security/uptime argument, which is *very* important but he specifically asked about performance issues.
just my 2 cents
"I don't need a compass to tell me which way the wind shines." - Mr. Furious, Mystery Men
Do the math on these numbers, 20K hits per day is 20000 / 24 hours / 60 minutes / 60 seconds which comes to .23 hits per second. An Atari 800 on a diesel generator could probably handle that. These statistics would indicate that Apache is completely lame compared to IIS. We use both here but for different purposes and happen to run both on DELL hardware. Apache and IIS are not going to be an order of magnitude different on the same hardware. Those numbers _must_ be wrong. Call DELL and ask them. They sell plenty of RedHat Linux systems (I have some of them) and they _like_ Linux and OpenSource. If those numbers _were_ true, DELL wouldn't have published it. Selling RedHat+Apache is money in their pocket saved from the extra expense of NT+IIS.
Apache may well ble slower than IIS (and probably is, as their main focus is portability and correctness, while many performance-tuning tricks must be applied manually), but IIS is by far slower than Tux as tested by Dell. Links:
Try updating web pages on IIS remotly. ftp your changes up, then go to the browser and test them. Where are my changes? Wait, must be the browser cache. Hit reload. Same- old copy, no changes. Its IIS- the damn server has the page cached. Hit reload a few more times. You then go through some sort of server side errors. Go get some more coffe and come back much, much later. At some point in the future, IIS will pick up your latest copy. Very annoying.
Tux is a red hat product. it's a multithreaded http server that does parts of its good deeds in kernel space. Watch out for apache 2.x on linux 2.4, though.. It'll achieve much of the same.. Tux just beat them to it.
Stop the brainwash
The specweb99 pages do not include a single apache setup. Furthermore, there IS something like a Red Hat 6.2 system. BTW - I bet apache 2.x with the 2.4 khttpd also would make a killing ;)
Stop the brainwash
Heh, that's highway robbery. I hope your point wasn't how easy/cheap it is to find MCSE's...
Free music from Jack Merlot.
A 486 running Apache can handle 100K hits per day. I can't imagine why they would say that.
Now, if they are talking about Apache and Perl vs. IIS and ASP then there are so many other issues to consider that the HTTP server probably doesn't play a major role.
Humph, 100k hits, that's kiddy stuff.
Vanguard
That which does not kill me only makes me whinier
On my home setup, my web server is currently a 300mhz Celeron/128 MB RAM/6 GB HDD running WIndows 2000 Pro and IIS.
And IIS cab barely handle the 30 hits a day I get! I can't imagine it being used in a commercial situation. I've had nothing but problems with it.
except hosting multiple sites on one IP address, using PHP(they DO have ASP for apache), restarting a hung service should your ISAPI code have a bug in it. and serveral other things, I've programmed for both webservers for a few years now, both compiled code(modules, ISAPI) and script(ASP, PHP, Perl)
> what is linux 6.2 why, just as the rumored 1 windows install that works (on a certain mr gate's desk) this is the secret kernel that linus keeps for himself.
Sitting Walrus Blog
I get a password dialog box.
120 day evals of NT and IIS have been free for a long time.
This is from Dell, who has close ties to MS, including the much litigated preferential treatment. Not to mention, Dell's whole Ebusiness endeavors run on IIS, and Windows platforms as a concession to MS. They are a "reference site" for Microsoft's Ebusiness solutions.
I agree about the advertising. But then again, I firmly believe that's where DELL does it's damage by making sure that the CIO of the company get's the add's and using those #'s, which we all know to be crap, in nice bold type, makes them believe that NT is the way to go. And that's why we, the people who do know the difference have slashdot. To get the hard evidence and make sure that the CIO does not get carried away with #'s.
Why do overlook and oversee mean opposite things?
This isn't really a reply to the troll itself, but I would just like to express my confusion at the message starting off saying "cock" and "fuck" bbut then suddenly it becomes "cack" and "fack" as some sort of psudo-censorship. If you're going to use something like f*ck instead of fuck, don't say "you F*cking fuckheads" because that just ruins the purpose.
----
ADVENTURERS! - ANTIHERO FOR HIRE - CARDMASTER CONFLICT
FreeBSD's SMP is stable, but nothing to write home about in terms of performance. FreeBSD has always been more about stability than speed however, so this isn't too much of a problem.
OpenBSD needs to go SMP soon. For two reasons: first, because this is a good time for Freenixes, and we're gaining a lot of ground... lack of SMP is a black eye for OpenBSD at a time where they might expand their market. Second, the longer the OS is developed, the harder it will be to do. SMP-capability is not something that can or should be "tacked on".
IMHO, if you're looking at running GNU/Linux on an eight-way box, you should really look at clustering two four-way boxen, or four two-way boxen. I feel this is not a limitation of Linux as much as a limitation of Intel's current architecture, and that more than four x86 CPUs on the same board will start to get starved for bandwidth. YMMV. If you insist on doing eight-way x86 Unix, spend the $80 and get a binary license for Solaris 8 x86.
---------///----------
All generalizations are false.
--
I like to watch.
---------///----------
All generalizations are false.
--
I like to watch.
Really knowing how NT works is truly a black art. However, because of the extremely limited about that is user-customizable in the NT kernel, this is a skill which isn't much in demand... as opposed to the UNIX world, where an extremely knowledgable admin really can make a difference.
---------///----------
All generalizations are false.
--
I like to watch.
>Linux 6.2 Apache based machine only supports 20k-100k
I guess I've been spoiled by using AOLserver. On an old, used $500 HP machine running AOLserver we can handle close to 250k hits a day, and that's with every page hitting the database.
Scott
EventNation.com
Actually anecdotal evidence is probably one of the things Microsoft is most worried about.
Anecdotal evidence was even mentioned in Microsoft's Linux Myths page - "Reality: Linux Needs Real World Proof Points Rather than Anecdotal Stories"
When marketing departments all around the world falsify statistics, and fund misleading "objective" tests, its not surprising that people rely on word of mouth reccomendations and real world experiences they can see and evaluate for themselves.
I'd personally trust stories and anecdotal evidence from people who have performed similar tasks to the ones I do, over biased benchmarks and marketing hype.
---
Silence is consent.
I'm only running Linux 2.2.17 and Apache 1.3.12. Didn't know Dell was selling 6.2...:))
Dell might have included some "speed enhancements".....It is open source ya know.
We must work in different worlds. In the Linux/Apache world I inhabit, serving a million hits a day on a single machine is no big deal. Up until early this year, I had a dual-processor 300MHz Linux box with 256M RAM that served 1.4 million, in addition to doing DNS and mail (over 30,000 messages per day). It's load average started going over 1 during peak time, so I moved it to an 850MHz Athlon.
The whole thing is a typo. Dell rates Windows and Linux the same for performance. Dell's editor and proofreader screwed up.
-------
-------
"don't smoke, don't drink, don't fuck
at least i can fucking think"
Minor Threat
Wow - you must have a *really* good ad! 20e6 ppl is far from 100k pages per day. Back to topic: say those 100k pages are served during one hour - it makes 100k/60/60 or 28 pages per second. You still can easily run quite heavy scripting.
_________________________
_________________________
Spelling and grammar mistakes left as an exercise for the reader.
It's still a restriction that keeps the GPL from being 'free for everyone' :) If the software cannot be used, unrestricted, in any circumstance, then it is not completely 'free'. :)
I think that is the point the person was trying to make. Whether or not it is an arbitrary restriction (which is subjective, anyway) is not the point. The point is that there is a restriction, so it is not free.
There is, of course, the argument that this restriction is what is needed, since it maintains the 'freedom' of the software for future generations. Again, this freedom is limited to those who license software in a way that is compatible with the GPL, not freedom for anyone.
So, it comes down to forced morality being used to keep the software available perpetually. To those who will play by the rules.
Oi, I need more caffeine.
Check the CPU count on those benchmarks. Are those really "virtually identical" configurations? (They may be, because I haven't read the fine print and may be misunderstanding something, but they don't appear identical to me.)
"Those who have never entered upon scientific pursuits know not a tithe of the poetry by which they are surrounded."
How would you pitch the fact that something will work soon to a PHB?
Im not saying it cant be done im just asking how do you get them to trust the whole open source process?
Jeremy
That feels like putting words in my mouth.
"Try that with xyz" I never said a thing about Microsoft or how it relates to the topic.
The fact is evne if you can test it out, when you need a solution you cant wait for a devleopment cycle to be completed. When I have a deadline or a timeline to work on im not going to say oh lets just wait for Linux to encorporate this feature when others operating systems already have it or can do feature xyz.
Its nice to know whats ocming and be able to tinker iwth it and paly with the source, and I stand by FreeBSD, Linux, and any OS with an open development cycle because of this access.
Plain and simple tho, if I need something now I cant wait for Linux to have it, evne if it isnt vaporware, its still "not usable ware"
Jeremy
It out performs apache on static file transfers. It's when you start doing serverside processing that Apaches strengths begin to show. Given that both will happily saturate a T1 anyway, I suggest you use other factors to decide
AGP stands for adcanced GRAPHIC bus. that's all it was designed for, and since almost all new vid cards are coming out on AGP, and the nature of the AGP bus means that there can only be one slot, there goes anything but vid cards in AGP. I haven't looked in any real large scale server boxes in awhile, is there any "upgrade" to PCI?
How many motherboards have onboard video AND an AGP slot? I've never seen one, can somebody give me a link to one that does?
Lets see I recentley load tested a couple of my front ends because I was worried what load they would be able to handle... FBSD 4.1 on dual PII 733's and 256mb RAM and with 10 machines hammering them for a 24 hour period I came up with an end result of 197 hits per second for one server... Multiply it out for one day (yes I did actually hammer them for a day) and you get 17020800. I beleive this is a bit better than the article states... Divide this by 10 to include images et all n you still have 1702080. You can determine the rest yourself.
-mutter- something something something...
It is old Linux 7.0 is out now!
joking
"`Ford, you're turning into a penguin. Stop it.'" -THHGTTG
"areas in which IIS simply falls on its face".
Nice talking, but any claims to proof that and which also state Apache doesn't fail in these areas? (you mentioned reliability too :))
IIS isn't insecure, nor unstable. The insecurities are in the dll's used to handle some requests. (like the .htr bug). The IIS process almost never requires a restart, only with certain (but already patched) issues. Apache leaves you with defunct processes from time to time, and an insecure mod compiled into apache will render apache insecure too.
Please, next time, come with hard proof (not slashdot stories, I ment proof, not dinnertable talk! ;) ), or include more details. IMHO your point is now a bit of a troll.
--
Never underestimate the relief of true separation of Religion and State.
You now have at least some redundancy.
The numbers are way out of whack though. It is *possible*, but it sounds like one of those tests where they take a SCSI card that has an early driver for Linux that doesn't even support disconnect/reconnect and compare the "same" performance with IIS. Where if you simply compared the highest peformance configurations (especially if they are the same cost including software), Apache should be VERY competetive.
Do you want MSW2K+IIS, or a few more processors and/or gigabytes of main memory?
For static pages I get about 800 hits per second for 50 concurrent users ( tested using apachebench). 800*86400=69 million hits/day.
;).
/home/apache/bin/ab -n 1000 -c 50 "http://localhost/"
This is ApacheBench, Version 1.3c apache-1.3
Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Copyright (c) 1998-1999 The Apache Group, http://www.apache.org/
Server Software: Apache/1.3.11
Server Hostname: localhost
Server Port: 80
Document Path: /
Document Length: 3777 bytes
Concurrency Level: 50
Time taken for tests: 1.243 seconds
Complete requests: 1000
Failed requests: 0
Total transferred: 4091087 bytes
HTML transferred: 3780777 bytes
Requests per second: 804.51
Transfer rate: 3291.30 kb/s received
Connnection Times (ms)
min avg max
Connect: 0 5 15
Processing: 16 54 55
Total: 16 59 70
This is on a Dell Poweredge 1300 PIII 500MHz 512MB 9GB.
Database fastcgi perl apps can do 3-70 hits/sec (very database dependent - bounded by database performance/tuning). Plain fastcgi perl can do about 200+ hits/sec.
The figures are so out, I wondering if there's a hint of troll somewhere
We've bought many Dell servers, mainly for Linux use and they are reliable- so far zero hardware problems. RH6.x Linux installs easily. New kernels work fine, only issue was the linux based megamgr for the AMIRAID didn't work. but it's fixed as of 2.2.16-3.
The poweredge 1300 made really good Linux boxes. So it'll be a shame if Dell screws up their own market. Well if they don't want the Linux market, Compaq and the rest are ramping up for it, so far I've installed RH Linux on a HP workgroup server with no probs either...
Cheerio,
Link.
that reported lower sales because he has missed the european clue train that folks want to buy athlons while he is trying to sell overpriced intel rot, or maybe the same michael dell that was one of the premier "bundlemeisters" for m$ personally, i would think a little about conflict of interest here... was the claim for apache under linux or apache unter nt? having some microsquish type configure the linux system so it swaps from the floppy can really do a job on your performance bigtime...
Allow me to quote an article on cnn.com concerning linux performance under heavy loads
S .idg/index.html
http://www.cnn.com/2000/TECH/computing/01/24/NO
"Linux did not perform well handling large loads - those tests in which there were more than 100 users. Under heavier user loads, Linux had a tendency to stop servicing file requests for a short period and then start up again."
Only the State obtains its revenue by coercion. - Murray Rothbard
Step 4: Show your boss that OpenBSD has no support for SMP.
Step 5: Show him traffic stats for cdrom.com
Only the State obtains its revenue by coercion. - Murray Rothbard
http://winbeta.microsoft.com
Only the State obtains its revenue by coercion. - Murray Rothbard
These companies that do these tests, they don't turn caching on in Apache, because it's not on by default, so static page performance suffers a lot.
---
GetSystemMetrics(SM_SECURE) == FALSE
From the little bit of info in this Ask Slashdot question, it is not clear why you are so focused on Dell. Have you already ruled out other vendors?
Consider VA Linux or Cobalt. I bet they edge out Dell on TCO for servers.
It seems that the only reason left to go with like Dell is to pay relatively less money for the standard MSFT desktop lock-in. Even there, I can say that their budget machines are a bit crappy--oh well, at least a good match for the software qualitywise.
And since you are considering a free software solution, you should pick a more neutral vendor. In any context in which MSFT and free software are being compared, Dell routinely implies that free software cannot hold a candle to MSFT. The only time Dell does not badmouth free software is when it thinks it cannot sell proprietary solutions.
In short, the best thing to do with the has-been DELL is to sell them short, just like everybody else on Wall Street is doing right now.
This should be mod'd as super-funny. I rarely laugh out loud at the printed word, but this did it for me today. Nice job :)
Praying for the end of your wide-awake nightmare.
Well I agree with you completely, but this is the kind of thing you must expect from a typical slashdoter (doesn't matter how ridiculous the story, as long as it's anti-m$)
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
100K per day sounds awful low to me. I've never run a big apache server off an intl/linux box, but the group I work for runs a small web farm on ultrasparc/solaris/apache boxes.
I'm looking at one of our solaris boxes, a U450 with 4 gigs of ram, and we've processed 2,048,607 raw requests (this includes graphics) today. Actual page hits, all of which have some dynamic content, comes out to 439,026. And there are 6 more hours left in the day.
A four processer machine with 4 gigs of memory is a big box, but it's not huge. I'm wondering what kind of processor/ram configuration the PC that DELL tested on was. I don't believe a comparable machine on an Intel/Linux box could only handle 100k a day.
Over 11 machines with similar configurations (a few only have 2 processors), we serve around 300,000,000 page hits a month. I find it hard to believe that Apache on Linux/Intel hardware would fare so much worse.
Jim
I understand that it is hoped that the advanced features of TUX 1.0 will eventually make their way into Apache.
I understand you didn't check your facts. The main feature of TUX will NEVER make it into Apache. It's impossible because the main feature of TUX is that it is compilied into the kernel. I'm not defending Microsoft on this one, I would consider uptime and security to be important factors, more so than performance. If they are looking for performance only, tell them to look at using Netscape iPlanet (sorry, but TUX isn't by any means ready for prime time) on their OS of choice.
Politics, Culture, Food?
touche
:-)
I bow to your superiour research
Politics, Culture, Food?
Zdnet Linux 2.2/Apache benchmark
Ludicrous TUX/2.4 numbers
Yes, 2.2 with Apache is a little slower than NT, but should still handle >500 hits/second. You'll run out of bandwidth first. Tux on 2.4 crushes all opposition. Static web-server performance is a useless pissing contest, and anyone who seriously worries about pure speed doesn't know what they're talking about.
With an average filesize of 10k, 200k hits/day is around
0.05MB/s. Not only is that not a network scalability limitation, it isn't very plausible.
Of course, alot of the action takes place in kernel space, since IIS uses IO completion ports and asynchronous IO.
My point was that any kind of real world usage pattern is very spiky, and you have to plan your capacity around peak usage, not average usage.
Why is it that the proponents of "one nation under God" are so eager to get rid of "liberty and justice for all"?
Errr, just for the record: an AC post starts out as zero, probably on the assumption that an AC better had something to say before relinquishing the minimal ID that slashdot users provide.
Bert Driehuis -- All I asked was a friggin' rotatin' chair. Throw me a bone here, people.
Sigh... If it make you feel any better, I have three Microsoft Certified Professional certificates, and I still don't like IIS much.
I'm an acknowledged elitist. If you have no clue about using gdb, or if you cannot find a web designer that is fluent in HTML rather than FrontPage, then by all means use IIS.
Performance is a very relative commodity. "You go to pieces so fast, people get hit by the shrapnel", as Zaphod Beeblebrox put it so eloquently. All IIS servers that I know that are reliable from the users perspective that I am aware of, either have a highly trained staff of NT gurus on hand, have 24x7 operators to hit the reset button when a blue screen occurs, or reboot the server nightly. All in all, a rare sight.
This is not my idea of reliability, but it is good enough for the vast majority of web sites.
Really, people who are not prepared to deal with the hardships of UNIX should probably run NT. I hate the would-be UNIX experts that drive down the UNIX mailing lists with questions that are answered in the README.
Now, that said, considering the level of expertise it takes to keep IIS up reliably, the whole easiness argument falls flat on its face. IIS is no easier to keep up than Apache. And Apache in its default configuration may have been overtaken by IIS in speed, it will survive if a stoopid programmer goofs up some CGI script. IIS (especially when used according to MS recommendations) will just give a 403 Server Too Busy, or worse.
Life ain't easy and there's no magic bullit.
Bert Driehuis -- All I asked was a friggin' rotatin' chair. Throw me a bone here, people.
It's a pretty well known fact that IIS kicks Apaches ass when only serving static content (dynamic content serving is pretty equal between the two IMO). If you're only serving static pages go with the Tux web server which just hit 1.0. It requires a Linux 2.4 kernel (since it uses the new kernel server) and it is in totally different speed league than either IIS or Apache.
Getting conclusive these days is hard. A lot of sites are becoming more intensely database driven. This means that when guaging performance, one must also take into account the database on the back end. I don't want to get into a flame war on which configuration rocks most, but there are some OBVIOUS things that should be avoided. I took a few statistics classes in college and have even done some benchmarking on my own (you know, to satisfy investors), and it is so easy to slant numbers your way. So I say, try a few configurations for your project and figure out what is or is not working. You'll eventually find something that works. By the way, we run DELL servers--some with Linux, some with NT. We don't get enough traffic to grade one better than the other, but have had problems with IIS conflicting with other software. No problems like that with Apache.
-- yawn. --
1) Dell printed an ad that said Linux/Apache tops out at 20k-100k requests per day.
2) Your CIO saw the ad.
3) Your CIO believes Dell without questioning or thinking about their claims.
The issue here isn't Linux vs. Win2k, nor is it Apache vs. IIS. Dell isn't the problem either. If Dell can make idiotic claims like this and people believe it, they win.
If I were GM and published an ad saying "don't buy that Ford car, it won't go faster than 20 mph" would anybody believe me? Or would common sense prevail?
Once people stop thinking, the marketdroids are in control. Let's not let that happen.
Say my Apache box serves 100k pages a day. That's over 4k per hour. Suppose my peak traffic is during the business day, OK, that's still not more than 12k per hour, or 200 a minute.
The truth is, Apache can serve over 200 _dynamic_ requests per _second_ on an SMP box. I've done it.
It's very possible somebody has a website that tops out at 20k per day. That site would be annoyingly slow under light traffic, and useless with heavy traffic. The problem also would have nothing to do with choice of web server, OS or even hardware (my '486 can do better).
I have Apache running on WinNT (don't ask) serving far more than that, on a fully dynamic, database-driven site. Yet the box shows only a slight load most of the time. It would do no worse running Linux. And BTW this is the public site for a Fortune 500 company.
Yet instead of laughing at this ad, as he should, your CIO wastes your time (and slashdot's time) with it. That's pathetic. I don't know whether to laugh or cry.
My point? Shame on Dell for posting results from 2 wildly different studies on the traffic any given hardware can support. Shame on Dell for trying to pose NT as the "high-traffic" solution and Red Hat as the low-traffic solution when you have the same hardware. If they start complaining about not selling any Linux servers, then ads like this will be to blame (I've seen the advertisement, and it's completely nuts).
Nuts. Shame on any CIO who believes it. The ad is no more or less than I'd expect from Dell's marketing department. They are desperately trying to hold their lead in hardware sales. Dell has made pacts with Intel and Microsoft and is very afraid of their market right now.
If the machine is really only supposed to be serving html, this may not be a deciding factor. If the machine also has to, say, serve up database requests or something, then you don't want the web server taking the OS with it if it goes belly-up. If web-serving is all it does, then it may not really matter.
Athy, athier, athiest.
This is certainly not off-topic. I don't know if anyone else looked at the Specweb report,
but _Dell_ submitted those results on their own machines. While it may be true that Apache on
RH 6.2 doesn't do as well as IIS, you can easily install a Linux 2.4 pre-release kernel on RH 7.0,
and install the TUX khttpd. It may still be beta quality, but IIS doesn't even hold a candle to TUX.
I believe specweb reported roughly a 400% performance increase over IIS.
Does anyone have a link to the actual specweb results?
Random mutating sig.
he means megabit, obviously. A gigabit ethernet card maxes out at 1000 megabit / sec, or 125 megabytes/sec, so obviously he is not exceeding the maximum data transfer rate of the pci bus.
---
Desperation is a stinky cologne
Maybe you would like it better if i said Apache has easier security and modularity. And with security, easier is better.
A year ago i did some testing using IIS/ASP for a project, one that also needed basic authentication over SSL. Guess what? Since IIS uses the OS model of security, i was opening up server privledges to every remote user, unless i set the extremely comprehensive permissions just right, or hired an NT admin to do it all the time. Even then, i can't trust MS to get it right. NT interfaces are just not geared towards easy viewing and updating of security permissions.
I was attracted to the ease of ASP development, but i didn't like being beholden to the windows NT platform for my business. Guess what i did? Found Apache::ASP and used the easy ASP object and scripting model. Plus, forget about COM, i just use Perl as the ASP scripting language and Perl Modules as the object model. I can scale that to any Unix box.
I can't imagine being able to bring this up to your boss (unless your company is a publisher of p0rn). P0rn sites require a very high level of performance, throughput and availability. They are, after all, a busy ecommerce site. This was the subject of a slashdot article a couple of months back. As you can see here apache is clearly what most p0rn sites have found to be the best solution. I find this very amusing, since you will never find this statistic on a DELL brochure. :)
Off topic? The above poster was hardly politic and polite in expressing his point of view, perhaps it was a bit of flame bait. But off topic? His point is about the most ON topic I can think of. Dell may be giving some tacit support to Linux, and there is certainly no love lost between Dell and Gates, but the fact of the matter is that Dell was a major player in the MS antitrust action specifically because their business and profits are so closely tied to MS that MS was able to literally dictate terms to them. It is not only on topic to note this fact, it is NECCESSARY to note this fact.
On the other hand, if I could moderate moderation I'd give a plus one funny to whoever moderated me down to 0 off topic.
Honest, I had a pretty good laugh over it.
Is it the OS or web serving software that is the key?
"You're fired" I Just picked up the phone and hired MCSE for $125/hr. There are 4 of them standing in line for the position. Have a nice day.
Marginally of topic perhaps, but does anyone know how to get perl running for IIS ?
--
If Google really cared they would fix Android Chrome to reflow text, instead of discriminating
They published the source. If you go here, you should see kernel updates and such, basically anything Dell sent you with the system plus anything new, and you can download the source if you so desire. I did not check to see whether there was a patch for your specific issue, but people's suggestions to check scsi settings seem valid.
Dell never shipped X4XX servers with AMI RAID, though they did with X3XX.
OK, so how does Apache compare to other servers?
Apache has been shown to be substantially faster than many other free servers. Although certain commercial servers have claimed to surpass Apache's speed (it has not been demonstrated that any of these "benchmarks" are a good way of measuring WWW server speed at any rate), we feel that it is better to have a mostly-fast free server than an extremely-fast server that costs thousands of dollars. Apache is run on sites that get millions of hits per day, and they have experienced no performance difficulties.
Sounds like Apache knows what up... Free, stable, and moderately fast or costly, unstable and very fast...
Yeah but are you recording hits _served_? Or are you reporting requests where 80% of clients abort waiting for the page to load because the machine is too busi eating requests. I don't know the internals of how apache schedules requests but from personal surfing expirience that appears to be a common effect.
KidSock
Is the question about Apache, or Linux? A lot of big name places use Apache as their webserver, but they don't run it on linux boxes. They run it on Solaris, Aix, or some other Unix that can handle big iron well. We all know that linux has a way to go before it can outperform other platforms, including NT, on large machines.
--Mando
$5000 for Microsoft software and $3000 for a PIII IDE 256MB system vs.
$8000 in Dual Xeon 1024MB 15,000rpm Ultra160 HD and $0 in free software.
A comparison of a $10,000 setup with a $5000 setup is not fair! All these comparisons ignore the fact that in real life the money for the Microsoft software would be coming out of the hardware budget.
Power tends to corrupt, and absolute power corrupts absolutely.
Back in 1994/5 I ran a popular server that got ~50k-80k hits/day. This was using the old Cern server on a box with four 30MHz Sparc processors, but it did lots of other stuff too. I would think more recent hardware would do somewhat better.
-Fzz
I'll probably get flamed by one side or another for saying this, but as much as I like Apache, IIS, espcially IIS 5 running on Win2k does outperform Apache by a long shot on static web content. Apache does have some capbabilities to outperform IIS on dynamic content, but comparing win2k IIS 5 on win2k to Apache this advantage is hardly there anymore. The truth is, the TCP/IP stack in Win2k kicks serious ass and for some reason outperforms Linux's current TCP/IP stack by a long shot. Apache has some serious advantages though, including better security than IIS and better stability overall. Basically, an analogy that fits: Apache; driving around the corner at a safe speed, with no chance of crashing or rolling over, and your pasengers remain securly in their seats. IIS; driving around the corner at a very fast unsafe speed, where two things can happen. You could possible make it around the corner quickly, passing up all traffic and getting to your destination a little sooner, or you could roll the car over, crashing, and your pasengers could loose their secure location in their seats, and fly out to their deaths. Personally, I believe the Apache way is the way to go, reliability and security is more important than speed. That is why I love Apache.
The rack mounted boxen you mention should cost much closer to 2k per, leaving you plenty of dough to pick up a pair of CyberIQ HyperFlow3 load balancers to more than even the odds. You could quite easily serve 2-4 million hits a day off a setup like that, and with the high availability offered by the hyperflow, you could get really darned close to 5 9s in uptime.
Think outside the... Hey, where'd the friggin' box go?
I have, in the real world, run a production web server handling 3+ mil hits a day.
.. and I wouldn't even think about it with IIS unless you have ZERO dyanmic content.
The server was a PIII 350, 512M ram, 20 gig IBM LVD drive. Kernel 2.0.36 (this was a while ago), Apache 1.3.x. The Apache was home-rolled with only the static modules we needed compiled in. We tweaked Apache a bit to allow > 256 childern. We also tweaked the kernel to allow more open files and other standard performance related tuning.
About 10-15% of our content was dynamic (perl), the rest static HTML and images.
The result: 0.50 to 1.00 load avg
Response time was also excellent. We had about 300 apache slots used at a given time, peaking at 400. I give that a definite thumbs up
BTW : Your argument sounds like the foolish arguments back in the early days of linux when all hardware support was compiled in via make flags : That was a BENEFIT because it gave a super ultra customized kernel that was super duper optimized! Of course now Linux operates almost entirely like NT/2000 with loaded devices, as it should so the song has changed : Loadable modules rule! Apache is moving to a thread architecture (2.0?) versus spawning a separate process because it is ABSURD to spawn a process for every single user, and when that happens the same people who are now carrying on about the super stability of Apache will be talking about it's superior multithreading architecture.
Actually, it's only absurd when you have extremely heavy-weight processes like NT/2000 does. Keep in mind that unix processes are actually light-weight (in comparison to NT); much more akin to what you think of as "threads" under NT in terms of performance.
In that light, Apache's operation as a pre-forking server is actually quite robust. Most NT admins don't understand this because they have been brainwashed by Redmond into thinking "processes bad, threads good."
They are identical machines, literally identical machines but the claim is that the Linux 6.2 Apache based machine only supports 20k-100k hits per day while the Windows 2000 IIS based machine (remember, same config) supports 500k-1M hits per day.
Looks like someone is confusing Redhat with Linux again... does anyone know about performance differences with Apache and different dists?
You could convince the school to support your lawsuit against Dell for false advertisement. Lying bastards.
http://en.wikipedia.org/wiki/2004_U.S._Election_c
Also, considering some of the huge sites running FreeBSD, I wonder how Windows/IIS would do against FreeBSD/apache? I'm not saying the free stuff has to win, I'm saying it has to be better, cheaper, and *faster* on a dollar for dollar basis.
I.e. if the Windows/IIS 1M hits solution was, for example, $50,000, but that was $25,000 of hardware plus $25,000 of software. Perhaps the free solution (FreeBSD/apache, yahoo & hotmail can't be wrong. :) would be $35,000 in hardware for the same 1M hits, but the software would be $0 so the "extra" couple of machines would be no big deal.
Hell I would rather have to manage a rack of 40 1U BSD boxes than 10 Windows servers! (considering the nightmare of remote access for just the first issue I have with Windows) So even if the free stuff was 25-50 percent of the speed of Windows, I would be willing to have the extra machines. :-)
I don't think Linux or BSD has to consistently beat Windows in squeezing every last drop of performance out of the box, sometimes it is just not going to happen. But Linux or BSD needs to be able to consistently provide better performance at a better price. If this means a slighter higher investment in hardware, this will almost always be offset by the $0 software investment. All of this without even considering the stability, manageability, remote administration, etc, etc that makes Linux or BSD so much more powerful!
This was a study that showed IIS was better in serving static web pages only. Add dynamic content and the advantage quiclkly goes to Apache. There were some articles about this test last Fall but I don't have the URLs handy.
linux 2.2.16 apache 1.3.12 100% dynbamic pages 600k pageviews/day (3 mln hits)balanced between 2 servers (2xZeon 500, 1 gig of ram, adaptec 2940) at about 12-30% load each.
I have just benchmarked my machine for a simple request here it goes: Front End. Hardware: 2xPentium III 1 gig of RAM Fast Wide Scsi disk mirrored Software: Redhat 6.0 kernel 2.2.16 Apache 1.3.12 Resin JSP server Back End Hardware: E420r 4x400Mhz CPU 2 gig of RAM SOftware: Solaris 7, Oracle8i Static "hello world" page: 722.54 rec/sec Dynamic "hello world" page: 543.48 rec/sec Dynamic "select sysdate from dual" page: 412.34 rec /sec
Finally someone hits the nail on the head! Too bad someone is not willing to do a http://Consumerreports.org test
"Where can I get -credible- data to prove which is better?" If you don't have access to the information that Apache kills IIS, how do you know that it is, in fact, better?"People need reset buttons"
Funny that the first reaction is to ask what was wrong with the tests. Are some of you so wrapped up in Linux superiority crap that you can't see how another OS might possible, just possibly outperform it in some areas? z
No, IE doesn't run in kernelspace. I was just using it as an example of a complex program. Recent versions of windows do use it as the primary interface shell though, which is almost as bad and why it crossed my mind at the time.
---
Where can the word be found, where can the word resound? Not here, there is not enough silence.
"Where shall the word be found, where will the word resound? Not here, there is not enough silence." -T.S. Eliot
20k-100k hits/day for Apache? Our machine, with a SLOW disk system, handles 100K hits/day with ease, and that includes the fact that over half of the processing time is taken up by the database server running on the same machine, which is about to be moved to it's own machine.
On a custom system using 10 StrongARM chips, Apache was able to serve out almost one million web pages **per second**.
steve
Oh, you're not stuck, you're just unable to let go of the onion rings.
My first hand exp with apache itself on SUN is great. I can tell you I watch apache handling literally a sustained 600 http processes and 80 raw http hits per second all day long (5 to 7 MILLION reqs per day). And it's only burning maybe 3-5% of a dual sparc 450.
So if this guy's software is "unoptimal for running on a unix-type system" he should re-write his company's project, to give a victory to *nix?
I was getting 8,192 SPECWeb on a G4 with MacHTTP. REally. Umm, No Joke...
Ask yourself, "what does a kernel have to do with a buffer overflow in a optional package such as apache, wu-ftpd or Pirhana(I know, not an exploit)?"
Nothing! Most exploits I have seen involved optional packages (wu-ftpd, apache, etc)
Anyone is free to create a Linux distro that is analagous to OpenBSD: it has no services enabled by default (save ssh), has a few "audited" packages, and tells you that you should use su. (I can hear the OBSD Zealots screaming now about how Linux will never ever ever never be secure like OpenBSD, because Theo tells us so.)
And why in the hell did this guy get modded up to +3??
Linux's security isn't so hot, and any commercial OS is better! (Holy shit,let's mod this guy up!)
Wow, what a brilliant comment!
I can't wait till the 2.4 kernel is old news. All the anti-linux "*nix users" posting FUD like this are going to be shutting up so they can go download ISO's from cdrom.com.
Linux is going to be one sweet kernel (it's not bad now), and I'm amazed at the progress the kernel developers are making.
www.kernelnotes.org
Why do I keep typing pythong?
TUX is being developed by RedHat. It will never make it into "Apache" proper, but will make it into an upcoming RedHat product, and source will be available.
When I hear the word 'innovation', I reach for my pistol.
I'm a consultant. My boss tells me what to code, and how to code it. If it's IIS, I tell him to shove it, and get another gig. I get my paychecks, big ones. I go home. I have the life I like.
If you are stupid enough to use slashdot stories to support any sort of business decision, your time would be better spent filling out job applications at McDonalds.
Conformity is the jailer of freedom and enemy of growth. -JFK
Linux is a kernel. [...]
- Yawn - pedantism is pretty boring. You know what I mean. The term "Linux" also means "the group of distributions that use the Linux kernel and include many open source packages".
I agree that the problem is in the open source packages, but so what? That's the place to blame in almost EVERY operating system, Windows included.
If you want it stated that "open source programs have nothing to brag about when it comes to security", then fine.
But none of this rhetorical nonsense matters one iota to the fact that the security under Linux operating systems has not been very good. Sorry to break it to you.
--
Sometimes it's best to just let stupid people be stupid.
Linux could have been securted if people wanted too. All you need is a few ipchains or iptables lines to make a box almost totally secured. The problem is that people leave their boxes connected to the Internet without configuring it.
Or maybe install a distro that is targeted on safetiness.
Why was this post moderated down? The AC had a damn valid point here, and a (Score: 5, Insightful) would be much more deserved than (Score: 0), because most people simply missed this and went straight for the flamefest...
I know slashdotters would try hard to get you some statistical data to prove (atleast for the time being) that Apache would outperform IIS. But what if Apache cant. Now.. i am not trying to start a flame war, I have used both, and not in small ECommerce sites, but I have used Apache in a site which gets millions of hits per day and the same goes for IIS. IIS is not as matured as Apache is, because Apache happens to be one of the longest and most successful Open Source project other than Linux.
IIS was pushed out of the door by Microsoft couple of years back when they didnt have any webserver to take on the Internet by its horns. So I wouldnt call it matured till 3.0. But 4.0 is a superior product and let me tell you, its only going to get better. Open Source or not, the ideas that would be implemented in both would be relatively the same. Security is one issue that both the webservers havent addressed to a lot extent and its where they needs improvement the most.
My question is when it comes to serving up Dynamic webpages the most efficient way possible, what matters is not whether its Open source or not, what matters is whether who serves up the most in less time, and what load it could handle. And if its Apache who comes out as a clear winner, then well and good, go for it. Just dont go for it because its the brethren for Linux and its Open Source. Businesses dont care less, as you pointed out in your post. They need superior performance and webservers that handle huge load and not something that bucks under pressure. I am sure there are reviews and benchmarks available to do that.
When all is said and done, I dont care whether the software that I recommend to my Clients is Open Source or not. What interests me is a 24x7 support, faster patches or fixes , fast deployment, security and performance. If the patch comes out of Redmond or a Basement in Illnois, I dont care.
Rapid Nirvana
that sucks, i'm not even getting the obligitory 20k hits that Dell says Apache "supports". Nothing worng yet, but, I always like to make technical decisions based on commercial propaganda, so looks like IIS for me.
-f
Umm, if you are looking to handle that much traffic I think you should probably look at something bigger than that Pentium you can pick up at the corner store. If you want performance look to Solaris and Sun's enterprise servers. If you are expecting 1M hits a day and you don't know where to look to find good information you should *NOT* be putting your site up. Sorry. I'm willing to drop my Open Source beliefs, but I would not drop to Windows. At 1M hits a day I think one should be looking at a Sun and load blancing.
Where can I get -credible- data to prove that Apache can outperform IIS?
NOT in an advertisment from Dell for starters...
find a really slow site (eg: pcworld.co.uk) page up netcraft and put the url in
It will come up NT/W2k/IIS guarenteed
opposite applies as well
Microsoft(tm) - a particular virulent virus that has infected most Pc's.
I've had a number of opportunites to work with both of them in a head-to-head environment. While the hardware config was never exactly the same (so benchmarks are fairly meaningless), the config was similar enough that I could make some educated comparisons.
For day-to-day content services, there was very little difference in performance. Raw HTML is a snap for both of these suckers. We never hit 1megahit per day so I never did get to see which one bogged down first, though.
IIS crashed a lot more often. We had an average uptime of about 3 weeks. Apache had an uptime of about 6 weeks, but bringing it backup was a commandline operation as opposed to a reboot-the-whole-damn-machine deal.
The userbase (internal) was much mroe comfroatble with IIS, and from an admin standpoint, it was easier for the IS staff. NT and IIS integrate easier with a wintel desktop environment, despite the existence of Samba.
All our business developers were easily able to transition to ASP, since all of them spoke VB. ASP performance was quite acceptable for the majority of our applications, many of which saw some fairly high (500khits/day) loads. However, the ASP.dll for IIS4 leaked memory like a sieve and was the primary cause of the uptime problems. Conversely, Apache provided a whole lot of technology options for app development, almost none of which were as easy to port business apps to because of the app's legacy. I'm just thanking god we never found a web-cobol product, or we'd probably have been stuck with that (the midrange guy kept trying to convince us to host our apps on the AS/400). It's a poor justification, but unfortunatley that kind of crap happens in the business world.
The overall system ran a LOT faster under apache, since it didn't need the overhead of everything else Windows decides to do...it was easy to strip the server down to basically apache and SSH. NT, well, you kill one service and god knows what else will stop working. Apache it was a lot easier to remote-configure the server since all you needed was an SSH client as opposed to MMC or the Gaping Security Risk That Is HTML Manager. And frankly, that affects performance too. If you have to log in on the console and run a Big Slow App in order to change a virtual directory URL, your server is going to feel that impact.
When it came down to it, IIS was an acceptable for an intranet server because of its integration and apps properties, and was able to serve apps and info really well for that sort of target audience. But the uptime problems made it a poor choice for high availabilty this-is-what-the-customer-sees kind of applications.
----
----
"I used to listen to Null Device before they sold out."
Regarding being only concerned about performance:
Performance indexes must be linked to cost. If it costs less to purchase and operate two Apache boxes and load-share them than it does one IIS box, and both configs perform equally well, which one do you think management would go for?
Assume for this problem that management is not focused on becoming an MS Partner, and that there are no restrictions on things like available rackspace.
Would this be like saying the air you breath is the same air those evil people with different values than you are surviving off as well so obviously a boycot is the best way to protest?
Man, if only people like you could follow through on your logic by testing your actions out, hell of a lot less morons running around.
Your argument holds if Apache has "slighly less raw speed". But 1/10 of the performance is not "slightly less" in my book. Todd.
Ok ok so its been so long since I"ve written a pthread that I forget the exact commands to start one. Anyway asside from the kernel space vs disk space issue, Apache is still forked and not threadded. Threads are sometimes described as lightweight processes, and as anyone who has played with threadded programming vs forked programming (just because they got a new smp box and were bored one night) will find that threads do indeed take FAR fewer resources than forks. I don't have my pthreads book with me here at work (Get the oreily book, the online faqs and tutorials don't quite cut it) I recall the explanation that performance would very from system to system, but 10 threads should use about the same resources as 6 forks in linux.
Of course the advantage to forking is that you can kill a rogue fork, a stuck fork, or a hacked fork in the case of a forked server, without any special tools other than su.
Still, I think a threadded version of apache would be good
You are only young once, but you can stay immature indefinitely.
I know! Throw up two boxes - one running IIS and one Linux/Apache and post two fake stories on Slashdot about SGI giving away free O2 Workstations or something. Link from the article to each box and see who gets Slashdotted first?
"On the Internet, everyone is an equal until they prove themselves to be a moron." - Emmanuel Goldstein
It's amazing how often people throw numbers around without even considering what they mean. 20k-100k per day means 0.23-1.16 hits per second. Even a 386 can handle that!
Nevertheless, PHBs will read this and say, "Ooooh, IIS' number is bigger. Bigger is better (I learned that in Business School)!"
Reminds me of the computer column in the Montreal Gazette a few years ago. The columnist's computer-illiterate friend running a small store was afraid she got ripped off by the computer vender, because her friend got a computer at the same price, and it had a "ram" inside, but hers only had a "mouse" inside...
Netperf works well for this sort of thing.
This sig for rent.
Tux is /not/ just for static pages. See previous articles on this where the author states this explicitly. It uses the kernel to serve up pages at a great rate of knots - how they are generated seemed not to have much effect in recent benches (although benches should never, ever be the sole criteria for setting up a web box)
-- Sig Sig Sputnik
You can split each virtual directory off as a separate process in NT 4, and with 2000 you have excellent granularity as far as chosing how to configure the server.
Inetinfo doesn't run as a kernel process, it runs as a user process in the system account. Secondly the service manager "requests" that inetinfo stop and any admin worth their salt knows that if the service is having troubles it won't respond to this friendly request and SCM has no other way of acting. In those cases you use your friendly "kill" command (in the resource kit or pskill from sysinternals.com) and either kill inetinfo.exe, or more acceptably you kill the out-of-process OLE object that your script created that is zombied and then shut down inetinfo.
It has been a long, long time since I've had to reboot a server because of a problem with IIS (i.e. there is no change to IIS that causes the server to have to be rebooted). Additionally a brain-dead IIS can be killed and then restarted. Of course on production servers the code is of a calibre that generally the service doesn't zombie, but developers have it happen a lot and know how to deal with it.
BTW : Your argument sounds like the foolish arguments back in the early days of linux when all hardware support was compiled in via make flags : That was a BENEFIT because it gave a super ultra customized kernel that was super duper optimized! Of course now Linux operates almost entirely like NT/2000 with loaded devices, as it should so the song has changed : Loadable modules rule! Apache is moving to a thread architecture (2.0?) versus spawning a separate process because it is ABSURD to spawn a process for every single user, and when that happens the same people who are now carrying on about the super stability of Apache will be talking about it's superior multithreading architecture.
It's interesting nonetheless.
You are probably a MCSE (Minesweeper Certified Solitaire Expert), aren't you?
Damn, you got me. Indeed I am. I am not averse to knowing what I'm doing, as apparently you are.
I'm sorry, but http://your.server.com/global.asa+.htr returning the plain text file is just dangerous, and if you've fixed that exploit, just check securityfocus.com for more.
Oh gee your case is proven. Portmapper for Redhat let a user root compromise the system (with a handy little root console and everything). I suppose therefore, by your reasoning, Linux isn't viable as a secure system? That's absurd logic my friend. Every major piece of software on the planet has things that don't quite work as expected. BTW : If "Chicago Title and Trust" used sa as their login for their database then they're idiots. In any case what exactly did you plan on doing with that password/username? The reality is that in the majority of cases there's shit all you could do as they usually don't leave port 1433 open.
If you weren't a M$ lemming on crack, you might realize that Apache is easier to program than IIS and that programming for COM blows. (Yeah, I've done the ISAPI filters, ATL COM objects and ASP.) ASP and other script in content languages suck. Sorry, but Apache and Tomcat blow away IIS and ASP/COM in development time and application stability/security.
I can see from your posting that you're clearly not an immature, zealot who couldn't hack the real world so I'm going to have to look into your claims so I can do some learnin'. Thank you for learnin' me good oh keemoosabi. Let me get back to Apache 1.32 on my FreeBSD box you moron.
+------------------------------------------------logged in as Administrator, not as a user in the admin group
As a point of interest there is nothing special about the Administrator accont. In fact it's common practice to completely neuter that account to remove a known for hax3rs. All that really matters (usually) is the "Administrators" local group, or Domain Admins which has priviledges usually only because it belongs in the Administrators local group (and it gets added to every machine that joins the domains local Administrators group).
However how exactly are you trying to kill the process? If you mean pulling up task manager then no, that won't work. Task manager tries to ensure stability of the system by not allowing certain operations to occur, such as the killing of processes running as services (there are some esoteric other conditions that are imposed). However to get around that as mentioned use the kill command (reskit) or pskill which do not try to be so polite.
To have the "user right" to use these tools against a process the user (or a group the user belongs to) needs to have the user right "Debug a process" or something along those lines (in NT 4 you have to show advanced rights I believe). Technically kill is opening a debug connection to the process and then destroying it.
Cheers!
Why not have a Windows/IIS expert and a Linux/Apache expert put forth their best web configurations?
Test how the web applications handle different loads, dynamic vs. static or whatever, then post the results as well as the configurations. This would allow people to see the configuration and results, then determine what to use and how to configure their web servers most efficiently?
All the money in the IT world, and yet we bicker on these small things. Could it be nobody has the balls to put forth what they consider to be the best they have to offer?
Despite a poor showing in the file benchmark, Windows 2000 came out on top in the long TCP transaction test. Windows 2000 is the only NOS with a multithreaded IP stack, which allows it to handle network requests with multiple processors. Novell and Red Hat say they are working on integrating this capability into their products.
CNN Article
however, most folks are going to be using the onboard video from the motherboard, and the agp slot (if there is one) will go unused.
knee-jerk? check. post? check. okay, time to read the article.
The only way to know for sure which is gonna be better for you application is to get test boxes and run them for a while... The tests that people like DELL run are usually fairly general in their coverage. If your gonna be using the server fro something specific then you're really gonna need to test the servers your self...
Your logic completely escapes me.
What does boycotting open source or Linux have to do with China's human rights?
But a company that helps to support China's poor human rights by doing business with them, and from the original poster's comment, making a "Deal" at that, just to line their own pockets -- that's a company that should be boycotted.
It seems to me that one can be concerned about human rights and be unhappy with some company cutting some kind of "Deal" (I don't know what it was) with China, just to enrich themselves, and not be in conflict. But what does boycotting open source or Linux have to do with it?
He uses Open Source and Linux because it is the best. The same reason China uses it -- which they have made clear -- it best addresses certian security concerns of thiers. The reason China uses Dell, is probably because Dell cut them some kind of "Deal". (Again, I know not of this "Deal".)
I'll see your senator, and I'll raise you two judges.
I don't presume to speak for anyone in the free software movement. That being said...
Just because we can't stop China from using free software doesn't mean we condone it. [Ignoring the issue of whether we should try to stop it.]
Where do you get the idea that by NOT boycotting Linux, I condone China's human rights. And you therefore conclude that various people who have replied to this thread are hypocrites because they don't boycott Linux? By my using Linux I'm somehow contributing to China's abuses or condoing them? By merely using Linux?
Anazing logic.
I'll see your senator, and I'll raise you two judges.
Just buy the 2000 version of the machine, do a couple of load tests on it yourself then reformat and install the linux/apache which you can get for free. Then if the linux machine outperforms keep it and return NT for a rebate.
As has often been said no benchmark published anywhere can apply to your system so its always best to measure these things yourself. The only downside here is the time its going to take but if thats not an issue then benchmarking for yourself is going to help. Also unless your site will be serving lots of pages then its just a matter of preference, whichever you prefer to use then go with it.
Also if you will be maintaining this then downtime comes into it, if you know linux then you can probably resolve any issues very quickly, however just changing over to linux without any prior experiance would be a move that could cause some grief. Its a great OS but like anything else it has a learning curve, you wont be a capable sysadmin overnight so make sure you have someone qualified to do all this work before making the move.
Slashdot: Proof that a million monkeys at a million typewriters can create a masterpiece
Just wanted to add to this conversation that I myself have done some testing. Might not be relevant, but here it is; Our website get 450k-500k hits a day. Our current system is a Sun Ultra 10 with Solaris 2.6 and Netscape Enterprise server. (Suitespot.) We tried to move the content to a dual P2 450 and NT+IIS. It croaked. Even after it had been tweaked by actual professionals, it BSOD'ed or just crashed after a fairly short time. We went back to the Sun box. (Customers were getting pissy about getting time-outs an' such... For our next system: OpenBSD 2.7, Apache, mySql and PHP3. Now THAT's gonna rock!
For most Internet related business, hardware is a minor cost. The major expenses associated with hardware are downtime when it breaks and people to run it. If you want to win this one, show that Apache has less downtime. Then show that your administrative staff knows Linux and doesn't know NT. Then show that the people developing it prefer Perl & friends to ASP & company. If you can't show these things, then you probably should be running IIS.
If you are modding me down because you disagree with me, use the "Flamebait" category, not the "Troll" one.
I have a compromise. Use Apache and Windows 2000, like I do. I'm serving a simple website for my honors society, and an eBook collection. Unfortunetly I don't have another computer so I have to host on mine. I like Linux and would probably use it for most things if I didn't need DVD and TV out of my computer. But Apache Win32 works well enough for me, while IIS was a memory hog, and was a pain in the butt.
If you had the choice of a webserver that could serve N hits a day, lacked useful remote admin capabilities, was filled with obscure nasty security flaws which generally couldn't be worked around, ran on an OS that requires 24/7 *in person* baby sitting, and one which served N/2 or N/4 hits a day on the same hardware, but which didn't have any of these problems, which would you choose?
Now if you have to spend another $5,000 in order to make the second webserver go as fast, but could then not spend 30k/year x 3 for people to sit in front of the machine in case it needed rebooting (a real scenario), how could you make a sensible business case purely based upon 'speed'.
I'm surprised the assembly OS weenies haven't jumped in with their own thread.
B>
Do you use GPL'd software to paint your house green? Do you use it to pick daisies by the side of the road at four in the morning? No.
A drivers license doesn't say you can't watch reruns of 'BJ and the Bear', but it might say you can't watch them *WHILE YOU ARE DRIVING* (Most states have laws about drivers not being allowed to participate in distracting activities)
This is the point. The software is being used for a purpose, and that purpose is evil (by most people's yardsticks). The software most certainly has a right to restrict it's usage in evil ways. I've read licenses which prohibit the use of software in nuclear weapons research and the like. It's not that uncommon.
To put it another way. Suppose you invent and patent a widget that someone else wants to put into a baby mulching device. Do you freely license the patent to him on the grounds that "Well, what he does with the patent is irrelevant to the patent itself?" I doubt it.
Now, how is this thread on-topic? It discusses IIS and Apache performance based on Dell's published specs. Certainly, the credibility of Dell, based on it's business practices, is on-topic to such a discussion.
If you need web hosting, you could do worse than here
Adding more memory does not necessarily increase performance. This is only the case if the tasks are memory bound and are using swap heavily.
If you need web hosting, you could do worse than here
Easy. Linux can be (and probably is) used to manage activities which involve violating human rights. Suppose a Linux system were used to manage troops in Tibet, or perhaps it were used to control Nuclear Missle guidance systems. The open source software would be aiding and abetting human rights violations.
While a human rights clause probably would not stop those people from using it (it would be under their legal control), it at least would make an effort to state that it shouldn't be used as such.
In any event. My argument is that if you're going to be on a high horse about Human Rights, start at home. Look at what your favorite OS could and probably is being used for.
If you need web hosting, you could do worse than here
As I thought. Human Rights are not as important to most people as Free Software. I'm sure the Tibetans applaud your morals.
If you need web hosting, you could do worse than here
Anyone that's ever implemented a caching scheme knows that there are diminishing returns on disk caching. As the cache grows larger, it takes longer to find entries in the cache, and the more housekeeping and CPU time is required to maintain it (remember, under SMP loads, you need to check cache coherancy and synchronize access to this as well).
If you need web hosting, you could do worse than here
Hmm... If doing business with China should horrify you, then why are you using Linux?
Remember, China has chosen Linux for primary use. If you are so concerned about human rights, you should boycott Linux (and all GPL'd software really) until the license has a human rights clause in it.
Are you going to do that? Or are you just blowing smoke about how important Human Rights are to you?
If you need web hosting, you could do worse than here
Well, you obviously haven't read the GPL then.
The GPL already *DOES* put restrictions on who can use GPL'd software. For instance, I can't use GPL'd software in my own code unless I also release my code under GPL. That means I am restricted from using GPL'd code in a non-GPL'd program.
The GPL already says who can and cannot us it's software and for what purposes. That means that the GPL already *IS* 'Free-only-for-those-we-want-it-to-be-free-for Software'.
If you need web hosting, you could do worse than here
Clearly you did not read the first few messages in this thread.
I said boycott Linux until a human rights clause is added to the GPL. Such a clause would legally prohibit anyone from using GPL'd software to violate human rights (of course it would still happen, but so does piracy. That doesn't mean we shouldn't use Licenses anymore because we can't stop it). Not having such a clause in the GPL admits that the software can be used for any purpose, including violating human rights and turns a blind eye to it. This is "silent approval", or knowing that something is happening and not specifically disapproving of it.
I'm actually not suggesting any such boycott. But what I am saying is that if you're going to suggest horror over a company doing business with China, then you should also express horror that your pet operating system is also used in such a fashion (for all we know, Dell is selling the Chinese computers with Linux on them).
Is it really so hard to fathom that I'm just asking people to be consistent? i don't care which way you go, but if you're going to criticize one and not the other, then that is hypocricy.
If you need web hosting, you could do worse than here
Of course any technology can be used for good or evil, but does that mean you have to KNOWINGLY contribute to evil?
The Chinese government has already admitted it uses Linux (and extensively at that). This has been muched hyped by the community as a good thing. If you contribute to Linux, your work is being used to violate others human rights. If you are a supporter of Linux, then you are supporting the "silent approval" of the use of Linux to violate human rights.
Hey, personally, I could care less. I just hate hypocrites. And people that bitch about how "horrified" people should be that Dell sells it's computers to china, should also be horrified that the Linux community and the FSF silently condone the violation of human rights by allowing their work to be used for such. If you're horrified by one, but not the other, you are a hypocrite.
If you need web hosting, you could do worse than here
The service pack is one conglomerated patch dealing with all the recent updates to everything Win2K. Get the service pack, and everything in Win2K is updated to whenever the SP was released. And of course you have to reboot after a service pack change! Sometimes, they update something in the kernel.
Here's one thing that Linux has never had, ever since its inception: dynamic swap file settings. Of course, this is because of the nature of Linux, since the swap file is really a partition. But it's fun running 7 copies of Quake 3 Arena and using up 1.2GB of RAM (320MB of that physical). Ever tried that in Linux? Hell no.
Don't forget, Dell hates it when you erase the operating system; if I'm right, the warranty will be void unless the installed OS was offered with the model. Ask them about these specifics.
And yes, it is funny how a subject can change so weirdly. After all, you're the one who forced me to start talking about Win2k.
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
Remember, a web server is only as good as the software it's running. Also, configuration makes a big difference, too. With only a few tweaks, most (if not all) of the security holes in IIS are covered. As for Apache, you always have to upgrade the entire server (can you say downtime?), and maybe even the Linux/BSD kernel (can you say MORE downtime?).
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
The documentation at the Dell server site contradicts this. For two equally configured machines running 2000/IIS and Redhat 6.2/Apache, Dell states that both will support 1-2 million hits/day.
I don't know if an apache is slower then IIS. But if it is there is a lot more you can do to speed up Apache and linux then you can for Win2000. First recompile your kernel and trim it down to a lean mean web server and don't forget to to put in a little khttp in there. Then edit your init scripts and take out every thing you don't need. If you really want every last bit of speed recompile your kernel, all your libs and Apache with optimized code for your CPU(s). And don't forget to set up apache for full performance.
I think that should give a apache that little extra boost.
When I was a boy the goverment stole everything from us.
There might be open source drivers for the add-in card you got on the 2450. The shoddy closed-source driver is for the on-board Adaptec raid card. Since the AMI is a great deal more expensive than the proprietary on-board controller, which do you think a customer will opt for? So Dell basically says "If you want an open source driver, you need to pay an extra thousand bucks." (And that's making the huge assumption that you are talking to a salesperson who knows how to do more than just take credit card numbers.)
Who cares how many hits a web site can take if the software it uses is unstable and prone to hackers?
IIS does not run in ring 0. Inetinfo.exe, the process that hosts the various services that make up IIS, runs in ring 3 (aka user mode).
It certainly calls functions in ring 0, as does every Windows app, but it does not live in ring 0. Check out Inside Windows 2000 by David Solomon and Mark Russinovitch for the details on what does and does not happen in ring 0.
Jeebus
How does it not? Quake pushes your machine to its limits. So if a app that pushes your machine to its limits runs better on M$ than linux then logically any app should run better. Including webservers you silly linux zeolots. --t0q
Yes yet another comment from that wonderful NT admin t0qer.
I love windows. If it wasn't for all those crappy winboxes breaking down al the time i'd be outta a job. Thats another story in itself.
Since the release of windows 2000 i've seen alot of people who once said "Oh big meanie bill M$ makes a crappy o/s" change their tune to "OMFG M$ finally got it right with a OS"
Yes folks win2k can smoke a similiarly configured linux box on web services. How do you ask? Well instead of boring you with hits per minute lets look at something we can all agree to. Quake 3 benchmarks.
Quake3 win2k runs circles around Q3 linux. So if it can do well there, it can do even better with web services, since it isn't drawing complex 32bit textured mapped 3d eye candy.
Yes I know you linux guys don't want to admit it, but thats a fact baby.
--t0qer
Okay, this is possible, if the PCI busses are seperate, at least 2X, then I'll buy it, but did you divide by ten or eight? Either way, it's somewhat silly.
Your post makes me laugh.
When they made you go out and pick up lunch, what did you get?
Buwahaha
Silly Rabbit: tricks are for kids.
You idiot. China doesn't make and sell the most distributions of Linux now, do they?
"My job is being right when other people are wrong." -- George Bernard Shaw
In my opinion this test has is similar to the one Mindcraft had. With the SMP patches from BSDi FreeBSD is going is going to be an amazingly scalable operating system. Does anyone know why they did not include FreeBSD in the SpecWEB 99 test? I think a neat test would be to network two of the exact same machines together with a ethernet or gigabit card, one running FreeBSD or Linux and the other running Windows 2000 server. Then have them simultaneously ping flood each other while fetching files via ftp and http and port scanning each other. Have the FreeBSD or Linux machine running KDE or GNOME with a secretary running WordPerfect and typing a letter dictating by her boss while she does not even know that the machine is doing all this stuff on it to have it compare equally to the Windows 2000 GUI with the same office setup. Then see who crashes!
Alot of this has already been said, but here's my take on it..
1stly dont assume that DELL have got this wrong, particuarly if there using fairly 'out the box' setting sfor linux and Apache. IIS is designed to run on the sort of hardware that DELL puts out, were as Apache is designed to run on pretty much anything (much the same could be said for NT and Linux/BSD)
This means that alot of work will have to go into optomising Linux and Apache for the open source solution to perform optimally. [for instance its typical to get a 2-fold or higher increase in disk IO by configureing hdparm etc correctly] It may be that DELL havent put this effort in, perhaps because they recieve less mark-up (and tech-support revenue??) or because they have less linux expertise. Its also perfectly possible that they have, and that, on their hardware, IIS is faster
It might well be worth contacting DELL and asking for more details on the setting used in the test, perhaps even trying to talk to someone involved in doing the testing, to see how much optimisation was done. Even if the Apache version wasn't highly optomised, you have to consider wether your company has the skills to optomise [and support you optimisations] yourself. If not you might well be better going for IIS.
You should also point out the importance of stability for the server, particularly if its for e-commerce, and ensure this factor is taken into consideration.
It may be that NT/IIS is the best solution for our company, dont worry about this, given its market share, and the fact that its free anyway, your usage is not going to have a great effect on the future of Apache. ☺
Why are you buying servers expecting the vendor to do the work? I buy Dell servers, I put the OS on them, I support it. When the HW breaks I call Dell and they fix it. Why do I like Dell? They always fix the HW fast. The OS is my problem. BTW I have a 2400 that runs RedHat6.1 fine. No complaints. HasH_Browns
scattered covered smothered chunked
Bingo! I'd moderate this up if I could. I'm of the same sentiment. I'm a developer. My boss tells me what to code, and how to code it. I do it. I get my paycheck. I go home. I have a life.
That's just silly.
And unprofessional.
I never saw the ratings for the Aussie Olympics... didn't something like 10 people watch it?
And that's my $0.32 (adjusted for inflation).
Great analogy- so your point is that school is not the place to learn how to read? Did the people you learned from treat you like you treated him? You have an abysmally low EQ.
That's not what you're searching for... the Linux machine there has 8 CPUs while windoz one just has 2. But look at the second quarter benchmarks. On two machines with the same configuration (2400/667 & 6400/700) Tux/Linux doubles IIS/windoz performance!!
Show you boss this if he is looking for performance http://www.oracle.com/guarantee/index.html?content .html.
Oracle claims they have it and are offering one million beautiful dollars if
they don't. They say that they can beat IBM and Microsoft solutions hands down.
BTW their app server runs Apache for http.
Crack |
I personally like Penguin Computing, cheaper than VA or IBM... still, all three are good choices...
Shit adds up at the bottom...
Even though I think Apache is faster in most cases, performance isn't the only point. ASP is nice sometimes, but it's a pain in the #*% to administer most of the time.
I would have to say that you should not trust anything that Dell says. Furthermore you should never ever under any circumstance buy any computer that is allready assembled and has software pre-installed if you are fully capable of building it yourself. Here is a perfect example of me and my roomate's computers. He has a brand new dell dimension 4100, it is 800mhz P3. I have my P3 450 I built last year. My computer never crashes unless I run unreliable software, like Netscape 6 or a test of some game. His computer fails often. I installed Red Hat 7 on my computer with no problem. I installed it on his computer, and it worked, but the Dell BIOS always starts windows despite LILO. So to get to linux he must use a boot disk. ALso his Linux failed the other day and had to be re-installed. And to top it off I ran 3Dmark 2000 on both machines. I even closed all the programs dell threw on there that eat up system resources. I got a score of 1100+ he got a score of 800 something. If you want a web server just buy a big case, dual processor motherboard, two chips, some RAM, a RAID controller, a bunch of 10000 rpm SCSI drives, download red hat 7 for free, and of course a video card and a sound card, which really don't matter in a server that much. Even though the money saved by building computers is not as significant as it used to be, when compared to Dell you always win because they charge 1000 dollars for putting there name on the case.
The GeekNights podcast is going strong. Listen!
If money is not issue there are lots of better solutions to run web server business than yours W2000 whatever.
Eg. S/390, IBM RS/6000 S80, Sun E10000..
Have you ever heard term "server consolidation"?
"Certified hardware". I hope that S/390, S80 or E10000 never get certified by MS.. :-)
"I'm just putting my nose in thing that i really don't know anything"
Pulpero, Nitwit!
And alway remember, Stupid American,
CHiCoMM phUcK1ng 0wNz j00!
And alway remember, Stupid American,
CHiCoMM phUcK1ng 0wNz j00!
-- The Yellow Fellows
Well, it's always the best to see it yourself. My machine (Slack 7, Apache 1.3.x) is able to serve about 100 requests per second. That means, it could serve about 8.640.000 requests daily. And to spice it up, my machine is running on a 500Mhz Intel Celeron with 256MB RAM. Oh, and I've run this test on a PHP file, so on a flat html file it could get a bit better. Don't belive the ads.
Didn't a conclusive report come out that stated that Apache worked better on clusters of one processor computers, but IIS was faster on a single SMP machine?
I've got a Dell 2400 here at work running Red Hat 6.2 with some strange proprietary, closed-source drivers for the RAID array. (They stopped shipping these with megaraid.) Under heavy loads, they pause every twenty seconds or so for about five seconds. I have no idea what they're doing, but that would definitely slow things down.
Stuff like that could contribute to low performance.
I got my Linux laptop at System76.
I seem to recall some kind of web-serving performance survey from a while back in which FreeBSD kicked Linux's ass (along with that of NT). Could this be an issue of the TCP stack in the 2.2 kernel series?
I know your joking. You have to be. This post is so backward you could'nt be serious.
You're joking, right? Hosting multiple sites on one IP address is horribly easy. Look up "host headers" and you'll see. Using PHP? you're joking here again, right? Add the file association for the PHP interpreter, and off you go! In addition, if someone were to take the time to write a PHP engine that integrated with the WSH (someone may have already done it, I'm not sure) you could use PHP right in ASP pages. ActiveState has already done this for Perl. Someone else already covered killing hung services, I'll let you re-read their post. When you said you programmed for both web servers, you meant years in "internet time" (IE, 2 days) right?
Knowing them, they turned on host name resolution :)
dellhost.com is primarily dell linux servers, hosted at interliant. Wonder why they'd want to shoot themselves in the foot on this topic?
===sam=== free nessus vulnerability scan = www.vulnerabilities.org
When setting up a server, you have to ask yourself several important things. How much do you want to spend? Which operating system and software do you know better? (If neither, Microsoft has a much smaller learning curve) What exactly is it going to be used for? If you are serving up webpages, how much bandwidth do you have? How many people will be accessing the server at one time? Is it going to be used in development? How much does remote administration mean? How stable is the distro you wish to use? Do you wish to use a distro or build from scratch? Linux is really nice because it is really cheap. If you want to modify your setup, you can change it via telnet. Setup changes don't require reboots as often. Windows and IIS are really nice do to the ease of use. They do have a remote administration for IIS, but it's security and ease of use leave a little something to be desired. Therefore if you wish to make a change you have to be at the box. And I don't care if I get flamed. If you know what you are doing, Windows NT (sp5 and up) is just as stable as any linux box. Who will be creating the webpages? Do the devolpers wish to use Frontpage, Active X, ASP? As far as speed, that is determined by many factors. What type of content is being distributed? How well is it written? Is the server just going to be distributing homepages? Or is it going to be a frontend for a server app? As many people have said, Apache wasn't built for speed. But neither was IIS. IIS is more integrated into the OS which is why it is faster. You should ask yourself... Should I try to push to get linux? OR find out which one would suit your (and your users) needs.
In the words of Richard Nixon, "I am not a crook."
Ok, I've read about every post and reply in this topic. So let's review:
1) All M$ enthusiasts bash linux
2) All Linux enthusiasts bash NT (but with accurate data)
ok, some facts:
-IIS 5.0 does outperform Apache w/Linux 2.2 kernel on static content ONLY
-1:10 ratio is obsurd, I'd be surprised if it was 1:2
-Kernel 2.4.0-test* does speed performance. This is due to optimizations in process cloning probably.
-Who uses only static content anyhow? not even microsoft does that. PHP? ASP? nope, not static. CGI? Nope, apache wins again.
-What is Dell defining as a hit? a page load? or a user session? 1 mil page loads isn't too bad, but a session where 10 or 20 pages are accessed by a single user is significant.
-Personal Experience: Running an Athlon 700 w/256MB of RAM, 2 10k RPM 18GB HDs, linux 2.4.0-test7 and Apache 1.3.12 + OpenSSL + PHP 4.0.2 -- averaging about 100k page loads per day and it uses only about 10% of my CPU, and still have 150MB of RAM free. This mean that 90% of my CPU is idle. Hmm, seems to me that 1M page loads wouldn't kill me. But remember I'm using PHP, so my content is dynamic and will take longer than a static page load. Funny....
-If you're only interested in static page loads, then use khttpd. geez, a kernel based web server, so-what if apache isn't the way to go for you...that doesn't mean that LINUX isn't!
-VERY IMPORTANT!!!! QUAKE 3 PERFORMANCE HAS NOTHING TO DO WITH LOADING OF STATIC WEB CONTENT. JACKASS. who actually hired that guy anyhow?
Dunno, just got pissed...had to reply...I'm sure someone will blow this off like I did everyone else's.
So, to translate to a language the boss will understand:
You've got 'Open Source' development tasks to perform, and if the boss orders an NT box, you won't have nearly as much code to upload to Freshmeat each week. Heaven help us, the server might be easy enough to administer that an IT flunky can keep it going. That wouldn't be good for Open Source because then your boss wouldn't pay you to develop non-business-related code to give away.
Explain this and watch your boss jump in panic at the thought he might get flamed by Richard Stallman.
Your CIO is in charge of what happens with your company's technology. What he wants will eventually happen, whether you like it or not.
The possibilities I see are:
(1) He's interested in making a thought-out and justifiable decision. Ads like this are evidence against your proposed solution. It's up to you to get evidence for it - evidence your CIO could show to his boss to justify having done it your way.
(2) He's one of those gullible management types who believes anything he reads - the Microsoft solution is presupposed, regardless of its demonstrable merits. The ad is just a way of getting you to shut up. In this case, nothing you do or say will make a difference - I'd look for another job, and it's the current company's loss when you find one.
Erm, no, he was talking about using system ram to cache the hard drive, not L2 to cache the system ram.
Dunno what world YOU live in, but if you're a unix guy and you get fired for being a unix guy, then:
(a) the place that fired you almost certainly wasn't a good place to work in the first place,
and
(b) you can get a much better job in no time flat.
Mark my words, in today's labor market someone who gets fired for being a unix guy is much better off for it.
- Duff, a Unix guy.
To illustrate: I have a pentium200 w/32 Megs of RAM runnign Slackware Linux 7.1 and Apache 1.3.12 compiled from scratch, and connected to the new with a T-1.
Using the provided tool "ab" (Apache Bench) I have concluded (don't flame me, I realize that benchmarks are useless) that my piddly little P-200 could sustain traffic of 25 hits per second. If that happened for an entire 24 hour period, I'd be looking at 2.1 million hits per day....which is complete bullshiz.
My point? Shame on Dell for posting results from 2 wildly different studies on the traffic any given hardware can support. Shame on Dell for trying to pose NT as the "high-traffic" solution and Red Hat as the low-traffic solution when you have the same hardware. If they start complaining about not selling any Linux servers, then ads like this will be to blame (I've seen the advertisement, and it's completely nuts).
Werd.
More specifically go to Third Quarter 2000 SPECweb99 Results
at http://www.zdnet.com/sp/stories/issue/0,4537,21961 15,00.html
complete with charts showing NT at the bottom.
I tried similar test with kernel 2.4.0-test8 and nearly got fired for shouting.
I was able to max out my switch and there is no upper limit in sight with the Cisco hardware I curently have available to me.
How did you manage to exceed the PCI bus speed? The fastest that I've ever seen was a GSN nic running on an RS6000 with a 64 bit, 50 MHz PCI bus, transmitting at about 300 MB/sec, and that was 100% limited by the PCI bus.
If performance is an issue, you should be looking at the "Tux" kernel-accelerated webserver. It absolutely creams IIS for delivery of static content.
--
Tired of FB/Google censorship? Visit UNCENSORED!
No, but there have been several other exploits discussed on Bugtraq this past week with OpenBSD that definately ARE root exploits (xlock, and format string errors in pw_error).
I just posted the above because I thought it was humerous. OpenBSD is not secure just because Theo tells you it is, or there is no open bug in an OpenBSD bug list.
And no, a DoS is not a root hack, but in this case, was easily preventable. The above occured because it slipped through OpenBSD's code review process, proving again that we are all human, that we all make mistakes, and that OpenBSD's open arrogance concerning their security doesn't amount to much of anything.
(People shouldn't throw stones at glass houses)
Besides...remember when it was discovered how volunerable NT was to teardrop attacks? This was also a DoS. It was due to poor coding on the part of the developers, and many of us in the security community held Windows very accountable to their stupidity.
Why should OpenBSD get any better treatment, especially when they claim to be so secure, and their code review so intensive?
Web Server Market Share
IIS 20%
Apache 60%
Hacked Web Server Market Share
IIS 60%
Apache 30%
Which implies that systems running IIS are 6 times more likely to be hacked
(Source - netcraft + attrition.org)
This is a characteristic of *all* software developers, be they RMS or Microsoft. Open source or closed-source. The difference with open-source is you can poke around in CVS and the mailing lists to find out how likely the marvellous new feature actually is, and how far away it is from appearing in a release that you'd actually consider using.
Any sufficiently advanced technology is indistinguishable from a rigged demo
--Andy Finkel (J. Klass?)
> Where can I get -credible- data to prove that Apache can outperform IIS?
:)
You can't. Apache's developers have never claimed apache is the fastest. One of them even replied to a tuning question with "...if by ``tuning'' you mean replacing apache with something that's actually fast."
Now Zeus is another matter entirely when it comes to speed. You can get a free demo of it. And its admin interface is nice and purty too
I've finally had it: until slashdot gets article moderation, I am not coming back.
Raw performance is only one variable in the equation. Sure IIS may perform slightly faster under certain circumstances, but at what cost? The instability and babysitting that come along with NT should also be a factor in this equation.
I could care less if Apache has slightly less raw speed; at least I don't have to hold its hand and be on call 24/7.
Basically it's freedom over servitude. Sure the work might get done a little quicker when the workers are being whipped, but I'd rather go a little slower and not be whipped.
Dom
"osm is an artist; I do not question his ways" --troll
Look at stability and total cost of ownership as well. In my experience NT/IIS go down far more often. Also if they cost you several hundred $'s on top of the price of the hardware you could run that much better hardware w/ Linux/Apache. Kernel versions, driver support for the chosen hardware, filesystems chosen, etc can all play a role in how many hits a machine can take. I'll tell you my favorite test.. pull the plug in the middle of heavy transactions from both machines and plug them back in and see which comes up quickest with least loss of data. Gotta love ReiserFS. :) You also have several alternative webserver choices for Linux is you want speed more than the other features of Apache. *shrugs* I'd never use NT for anything mission critical. To much experience with those issues. ;>
At what price learning? At what cost wisdom? The price is a man's peace of mind, and the cost is his life.
Adding more memory does not necessarily increase performance. This is only the case if the tasks are memory bound and are using swap heavily.
In case anyone out there read this and nodded and thought, "yup, ok, that sounds good," allow me to point out that this is 100% bullshit.
RAM is also used for buffers and cache, and both of those will make a large difference in performance. Almost all applications are memory-bound, and any application that is swapping heavily is on an overloaded machine and has nothing to do with real-world web service anyway.
In every case where you are actually using a server in a production system with more than a negligible load, adding RAM will increase performace - up to the point where your entire site is cached and all non-logging disk i/o stops.
You will see, obviously, a much greater increase in performance by adding RAM to a machine that is swapping.
Obtopic: No one can answer your question without knowing what kind of machine it is and what kind of data you're serving. At best, we can speculate.
See ya!
Blue
i browse at -1 because they're funnier than you are.
I don't want a lot, I just want it all!
Flame away, I have a hose!
Only 'flamers' flame!
Remember the day the satellite blew it's zap and all those pagers stopped working a few years back?
Well our company hosts www.panamsat.com the coporate site for the satellite company.
On that day the site was running along with 50 other shared hosting sites on a pentium 166mhz w/64mb RAM running Slackware and a 1.2.x Linux kernal.
Prior to that day the site would take ~1000 hits a day, but by the end of the day just that site alone took 1.2millon hits.
From that I know that a properly configured Linux server could take at least a few million hits a day.
Forget the FUD, if you want scalable stability don't use Windows.
Congratulations, you just figured out how to spend 2.5 times as much by replacing an experienced
Damnit, should have previewed... :-(
Congratulations, you just figured out how to spend 2.5 times as much by replacing an experienced < $100,000 salaried employee with a $250,000+ 'paper expert'. And make no mistake about it, you need someone full time to administer servers, especially ones that require as much tweaking and twiddling and handholding to keep running properly and reliably as NT/IIS does.
Vendor sponsored multiple guess 'certification' programs like MCSE are pretty nearly worthless for determining quality. While there may be a lot of MCSEs out there, how many of them really have good applied knowledge, experience and problem solving skills?
Microsoft is selling the PHB's a false economy with the way they try to make it out that any idiot out there can administer NT/W2K. And hiring consultants for this sort of thing is just plain foolish $ wise anyway.
I convinced my PHB's to switch to Apache very simply. I setup IIS and let it run on one of our installations. It ran fine for a few months, when nobody was using it much. Then the system finally went full time, and IIS started crashing every 3 hours. It couldn't handle the load (which was actually pretty minor). So I quickly configured Apache for NT, and set it up (after removing IIS) with the exact same dynamic content (CGIs mostly). It works without fail, and uses less processor time to do the same thing.
Now I'm using that as leverage to make them see why NT sucks so bad for this system. Maybe I can push them to a Linux or BSD type solution.
Sure, IIS may be a bit faster, but there's a huge difference between fast and stable.
---
- Give a man a fire and he's warm for a day, but set him on fire and he's warm for the rest of his life.
I know a admin for Expedia.com. He claims they their NT web servers are rebooted more than once per day. I also heard a talk from Steve Ballmer where he said that Microsoft.com was run on ~70 NT web servers by ~70 admins! He jokingly said that each admin must spend their entire day sitting in front of "their" web server and repeatedly ask themselves, "Did the server crash?! Did the server crash?!" If Microsoft can't keep NT up and running on their own Microsoft.com and Expedia.com web sites, how do they expect anyone else to do so?
I also know a web developer from go2net.com. They're a Linux shop, but acquired SiliconInvestor.com who used NT. They routinely had to reboot their inherited NT web servers every morning.
cpeterso
Oooops.
HEH i should have checked my arithmetic.
100k hits/day:
(gdb) print 100000.0/60/24/24
$1 = 2.8935185185185177
Urm
Thats 3 HITS A SECOND.
Time to fire your CIO.
I think my HP-200LX palmtop can serve that or better....
I threw together a mysql/perl/apache combo to replace a horrid sql server/NT app that sucked. With purely unoptimized sql and half-ass CGI.pm code, I got about 4 hits per second (which included the db lookups) versus 1 per sec on NT, with all of it running on a single cpu AMD K6-350.
I know, not definitive enough.... Don't forget to try ApacheBench for torturing the boxes. A company called Cyrano makes some good loadtesting software too.
See http://www.spec.org/osg/web99/results/res2000q3/ for a refutation of the Dell claims.
No, no, no. Internet Explorer is not in kernel space. Or at least it wasn't last time I heard, but who knows what MS has done recently. But that's not what everyone else is talking about.
We are comparing IIS - Internet Information Server - to Apache. They are web servers. IE is a web browser. IE is not the same as IIS.
IIS runs largely in kernel space for high speed on static pages, but takes a big hit on highly dynamic pages.
The new TUX web server for Linux is an optional kernel module which is kind of similar, but gives you the best of both worlds - kernel space httpd for very high speed on static pages, and it passes off requests for dynamic content to Apache which runs them very well.
Torrey Hoffman (Azog)
Torrey Hoffman (Azog)
"HTML needs a rant tag" - Alan Cox
IIS is known to be superior to Apache for *static content* -- but what about dynamically generated content, which is what most companies really care about *anyway*.
The kernel httpd will help for performance with static content, but it's not out of beta just yet...
My daytime employer handled 300,000,000 page views in August -- almost all dynamically generated, most on Apache/Linux, the rest on Apache/Solaris.
Our server farm is around 60-ish servers I think... That includes the two E4500's at the back-end. (one live, one hot-spare)
-JF
MrJoy.com -- Because coding is FUN!
Erm, don't mean to pick nits, but 100,000 into 86,400 seconds per day is one hit approximately every 1.16 seconds . . .
You should still fire the CIO, though.
They're taking their dog to get its two shots before it's too late. You're taking your dog there too, right?
Think about this.... Dell sells MS products and probably gets some kind of discount. Dell wants to sell Linux. MS gets pissed and says the only way their relationship will stay the same as it is now is if Dell gives better performance stats for MS products than Linux.
100,000 hits a day??? That's barely over 1 per second. I've done benchmarking on a perl based shopping utility running SSL on a dual pentium 233 (not PII!) under apache and it maxed out at about 27 hits/s. The perl code was terrible also. Whack some PHP on there and it would probably do over 100 hit/s sec no problem. I don't know where they came up with that 100,000 hits/day BS, but they're obviously taking 100,000 hits/day off the bong.
Need Free Juniper/NetScreen Support? JuniperForum
Back in the 1800s, Nietsche proposed the idea of an uebermensch who had the sensitivity and intelligence to rise above the daily petty lives of normal men. This entire "who has the fast web server" is one of those petty squables. The fact is that each is fastest in its own way. If you want to prove how good your OS/httpd compared to the other OS/httpd, then you'll certainly have enough statistics to back you up. So will the other side. The only way to win the war is not to fight it.
DELL was almost certainly quoting a specific test, possibly one tha compared dynamic content generation between Apache+CGI vs. IIS+ISAPI (CGI sucks vs. almost any alternative).
.. which is very exciting, but not as proven
as Apache. People not wanting to be early adaptors
may want to wait for a while.
I'd have to agree so much here. It's ridiculous how meaningless raw speed data is thrown about in this industry. For most sights, the ability to handle an astronomical number of hits doesn't really mean anything since they'll never see that sort of traffic. Other things are much more important.
First, ask your boss if your company realistically ever expects to get within more than a few percentage points of either supported hit rate. The idea here is that those numbers are meaningless.
Second, ask about failover support. With the cost of NT/IIS, you could afford another server to provide backup support.
Third, ask about reliability. The throughput of a computer is exactly 0 when it is rebooting, blue screening, or otherwise taking a break from reality.
Finally, if your boss insist that one small narrow view of the world is all that is necessary to make an important decision, then consider posting your resume on one of the many internet job sites. Your prospects look good in the current economy.
Aah, change is good. -- Rafiki
Yeah, but it ain't easy. -- Simba
In the benchmarks done about 6 months ago IIS did outperform apache, BUT that wasnt apache's fault. The fault was with the tcp/ip stack in linux, and it was determined that apache on *bsd would outperform IIS because of *bsd's better tcp/ip stack.
I like Apache a lot, but it is not buit for raw speed -- it is built for manageability. It's a joy to work with. The 100K hit figure seems a bit low if this is high end intel iron (it's only 1.15 hits per second); but even at face value, I'd expect this to go up considerably with the 2.4 kernel (multithreading); Squid can take up some of the slack too if you need to increase the static page performance several fold.
It's probable there's no technical barrier to what you want to do using Linux/Apache. However, what the Dell folks have done is throw a sales barrier in front of you.
If you are concerned about selling scalability to t PHB, why not AOLServer? It's open source, and AOL uses it to host their services to the tune of 28,000 hits/second or about a hundred million hits/hour. Granted this is not on anything resembling the iron you are buying, but if you're expecting to get into the rarified 1M hit/day range it will be nice to know that you have a place to go without rearchitecting when you are hit the wall. The money for a nice solaris box won't seem like much if you can avoid months of porting software.
Another thing to point out is that the max static pages benchmark is not a measure of speed, but of capacity. Within the reasonable range of the Apache performance, you won't see any difference in responsiveness. Thus if you are planning for less than 100K hits/day there's no reason to use this benchmark as a differentiator. Instead, focus on issues of manageability and TCO.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
You miss the point: What is Linux 6.2?
Hey, look, the Apache website server status says they're serving 17 hits per second. That's around 1.4 million per day. Looks pretty fast. Looks like they're running Apache. Hmmmm.
Hey it great to have so many choices.
A Dick and a Bush .. You know somebody's gonna get screwed.
War is necrophilia.
Having said that you could always use aolserver which is godawful fast and php right now and cook and go into overdrive when kernel 2.4 get out.
A Dick and a Bush .. You know somebody's gonna get screwed.
War is necrophilia.
The server is a Compaq Proliant with a 300 Mhz processor and 128 megs of RAM.
-Omar
PS -- The same box is also our primary SMTP server for ~500 email users (qmail, of course).
Try this one for more info.- --------
------------------------------------------
-------------------------------------------------
This sig could have been put to good use.
Ouch. I assume thats in addition to the shared memory (of course -- what else could it be? You say "per worker process"). So, I suppose you're looking looking at a 25 - 35 meg httpd daemon. That's pretty darned heavy
When you start using things like embPerl (a php-like extension to perl), and you have a large project (dozens of perl modules), all cached in memory, you can easily bring your perl-space alone to 6- 12 meg. With mod-perl off, Apache is like 2-3 meg per worker for me. With everything turned on, I usually get 10meg... If you don't properly manage your perl code, then things like database accesses can consume a meg or more for a single variable.. This is hard to de-allocate in such a way to free up the process's memory space.. That coupled with fragmentation can grow your memory size considerably. It's virtual memory, to be sure, but it still counts as disk swapping.
The size of the daemon totally depends on how much code is loaded at apache startup time.. If your
I suppose, though, that no-one thought to make perl interact well with copy-on-write memory semantics. It would be cool if it did.
Well, so long as you initialize _all_ your perl code in your httpd.conf or vhost files, then you'll get essentially that. Anything generated after the dameon spawns off workers is lost, needs to be reloaded / recompiled. In the first case, you take advantage of UNIX's copy-on-write. This means most of the constant data-sections (namely the in-memory text copy of the perl source code), won't be physically duplicated.
Unfortunately, unlike most programs, most of perl's memory is writable. Meaning, you have the core perl-interpreter which is static code, then you have the hundreds of K of perl-code-text, and their intermediate meta-data or compiled form. This compiled form does not get shared between forks, since the execution tree can change over-time. At the very least, it can't be registered as "shared memory", since it's possible for the contents to change. At best, you have multiple forked processes temporarily sharing a segment with the copy-on-write flag set. If ANYTHING in that 2, 4 or 8K page changes, you have to copy out the code.
Could you explain what you mean by "the cache lines are valid for all threads." Obviously, the code cache is going to be valid (just like in Unix, obviously), and the data cache is also going to be valid (unfortunitely, not like unix). But is the stack cache going to be valid?
Ok, thinking this over, I may not have all my facts straight. I know that the x86 uses a legacy addressing scheme, unlike most modern CPU's which have a flat virtual address space, the x86 uses segment selectors. This allows a process to utilize two segments that have the exact same virtual address but really be mapped to totally different physical addresses. A segment is an offset to the CPU address register which occurs prior to page-table mapping. It also provided an upper address bound to provide a first-tier memory-protection scheme.. It worked great in the 386 days, but in our UNIX flat-virtual-memory model it's not terribly useful. I think NT still uses it for some optimization tricks.
My problem is that I do not know exactly how Linux deals with this.. My assumption was that since each process got it's own segment selector, it would have a completely different virtual memory offset, and thus the addresses that the cache lines get are different for each process. From this, I assumed that two context switches would require cache-flushing. If anyone knows differently, please come forward. If this were true (as I assumed), then MT would have the added benifit of sharing a segment selector, and thus not requiring a cache flush.
As you can see, if my original assumption was correct, then none of the code, or data cache lines would be valid. If I was incorrect, then it's just like most any other processor. Both your code and data lines are valid for all MT context switches. It's true that you have independant stacks for threads, but they're in the same memory space, they shouldn't do too much harm to each other's cache. With MP, even here, you have the issue of starving neighboring process's data caches, since for for all copy-on-write operations, you have additional memory to cache and compete against. Still, unless my origional assumption was true, none of this really will affect much.
Also, what on earth can you possibly mean when you say "the apache processes run round robin". When did apache start including a task scheduler? Will it start coming with disk drivers soon, too?
Ok, I have made assumptions based on empirical evidence. The daemon process opens port 80 (or what-ever), THEN forks off a bunch of processes with port 80 still open.. This is a very bizzar situation. Normally you'd think of it as simply having the forked processing maintaining the file-handle from the original guy. For STDOUT, it's just the luck of the draw who prints out in which order when they're both performing prints. But for server TCP sockets, things get wierd. Lets say the central daemon never listens to sockets (
it's main role is to monitor the number of workers and fork off working copies as necessary), but all 5 of your workers ARE listening on that same socket. What happens when a browers establishes a connection on port 80? You have 5 guys listening and waiting to accept a new connection, which gets it? Well, it's the OS's job to decide, and evidence on my installtion suggests it's round robin.
Here's how I did my experiement. I used mod perl in order to have a persistent memory structure. I went to a web page that would print out to the screen it's process id (since it was mod-perl, it was the same ID as that of the worker). You'd think that you'd always use the first worker, and only use additional workers if the first was busy. But instead, I'd get a random worker. I don't recall if I worked out whether it was indeed round robin.
Think of it like MT with cond_wait and cond_signal.
All workers cond_wait on the port-connection. A seperate thread (the kernel in this case) finds a new connection to port 80, so it signals a cond_signal, and some undetermined guy wakes up first and establishes the connection.. The other guys go back to sleep until the next cond_signal. When the worker is done with the connection, it does another cond_wait, reentering itself to the queue.
The most simple implementation of such a conditional broadcasting structure would be to have a round-robin queue. cond_signal activates the guy at the head of the queue, and all people initiating cond_wait go to the end of the queue.
All of this is hidden by the socket "accept" function call, which should just be an interface into the inet kernel routines. I don't know how much of libnet is done in user space and how much in kernel space, but I'm sure that no actual MT programming is going on, since I don't see any IPC resources being consumed when children listen to the same socket.
To experiement on your own, just do the following (written in perl code for brevity).
#!/usr/bin/perl
use IO::Socket;
$sock = new IO::Socket::INET( Listen => 5, LocalPort => 5000, Reuse => 1 );
for( 0
fork;
}
while(1) {
$client = $sock->accept;
$client->print( "Server $$\n" );
$client->close();
}
Then repeatedly telnet to port 5000, and watch how it cycles.
I just tried it out, and it produces 4 processes. Telneting shows me that at least on my implementation with this few processes it is perfectly round-robin. This may not be the general case, but we can't deny that having 50 web servers is always going to spawn the first one (thereby reusing it's cach lines). If I am correct, then if you are underloaded, then having 50 apache processes will actually hurt your performance (especially if any of it goes out to disk).
-Michael
Hmm, I'd be curious to learn more about TUX. I was wondering what kind of performance you could get if you used an X-type event model using UNIX's select. Divide your apps into tiny chunks that take very little time each, give the apps priority queue's where tasks that take longer than one time-unit queue subsequent activities on lower-prio-q's. Then periodically poll for IO (for both disk-reads and Socket activity).
This is kind of like an RT system, where IO is garunteed to not be pending for longer than a certain amount of time. This allows thousands or millions of simultaneous connections with only the overhead of short-term context-information.
You get very fast connection-response time, and you could internally prioritize activities so as to minimize overloading (though this probably couldn't be true in general).
The bueaty of the system is that if you actually HAD multiple CPU's you could run multiple worker processes in a very similar fashion as Apache currently does. From this, there would be no incentive to run more workers than CPU's, since a given worker can handle an infinite number of connections (limited only by practicality).
This obviously wouldn't not support external API's such as CGI or mod-perl, though it MIGHT work with FastCGI or servlets, since all you need do is probagate IO.
-Michael
putting something complex like a webserver (or browser for that matter) into kernelspace is just asking for it
I assume you mean IE. I was under the impression that IE did NOT live in OS-space. I base this on the fact that when i?explorer dies, I can restart it without affect independant proceses (course, if anyone had a file-dialog box open, they'd freeze too).
Please correct me if I'm wrong, but Windows runs their utility DLL's in user-space. It's only the vxd's (for the win 9x world that is) that runs in kernel-space. Likewise NT 4.0 didn't let any utilities in that space. 4.0 allowed all things video to probagate, but that was it. No clue about 2k.
-Michael
I was assuming that much. We're talking about 500 THOUSAND connections though. I seriously doubt that a single connection is going to have 499,999 image links. :)
-Michael
-Michael
If it's static content, wouldn't TUX be the way to go? If course, I think TUX was tied to the 2.4 kernel, but I'm a little fuzzy...
-- You can't idiot-proof anything, because they're always coming out with better idiots.
In posing the question of where to get credible data to prove Apache can outperform the competition you are not being objective or serving your company well.
Apache will certainly win at some things but you need to look at the type of serving you need and evaluate which solution is best. That is not the same as going into an evaluation with a loaded question about proving your favored candidate can win. Typically you can find evidence and circumstances to support even a weak proposition. To do so really accomplishes nothing except to support an arbitrary agenda which is unrelated to your business needs.
So, figure out what you need and evaluate which solution delivers that at what cost, or at least figure out which solution delivers more of what you need.
If you're not prepared to do that then just pick your favorite, or roll a dice. Don't waste your time going through the motions.
Some points:
My list of good guys and bad guys:
Good Guys
- Linus, Alan and the rest of the community.
- SGI
- IBM
- Linux Vendors
- FreeBSD
- Matrox
Bad Guys- Micro$oft
- Dell
The above list is my opinion, feel free to draw up your own.Don't forget about Tux. If most of your content is static, Tux will be a very potent solution, IIRC. Since it runs in kernel space, you won't waste time doing context switching.
:)
Just thought, and it could be wrong...
jason
Have a good day?! Impossible! I'm at work!
It looks like Dell is speaking out of both sides of the mouth
Windows is the most popular OS for client PC's. I understand this, and accept the fact. How many end users have had exposure to other OS's? When people choose an OS, do they actually look at benchmarks, reviews, etc? We are talking about Web Servers, which careful planning is applied to. I have yet to meet a SysAdmin who chooses Apache or IIS because everyone else is. You have to do research on the matter. Which Web Server suits your needs the best? Which platform will you be running it on? Do you need enterprise support? What about modularity and scalability. What are you willing to spend? These subjects are addressed before a Web Server is decided upon.
Netcraft. They give some basic info about webservers; what people are using, what is used the most (*cough* Apache *cough*).
Dell would use Red Hat, because they are a Red Hat distributor.
The only way to get truly credible data would be to set up a test bet and test against what type of traffic YOU expect. General statements that Web server X is faster that web server Y are only useful if they are close to the type of traffic you expect. Propose to management that you try out both operating systems in an environment similar to production and see which does better. Much to everyone's disgust it may be that Windows 2000 may be a better performer for you but I think everyone involved would rather find that out before rather than after. Well that's my 2 Cents.
| Cunning Pike... Good Guy...
And if so, wasn't it basically a rehash of the Mindcraft benchmark? You know, the one that showed Linux slower than NT on a number of services, including web serving and file serving.
If so, this leads me to ask: are these Dells SMP? Because IIRC, the main showing of that benchmark was that Linux doesn't scale as well to multiple processors as NT does. And Linus even conceded this, and there are supposedly big SMP support improvements in 2.4.
If, on the other hand, the Dell boxes AREN'T SMP, I'm really curious where those figures came from. Because I don't recall NT being 2-5 times better until scaled WAY up on multiple processors. Maybe Dell gets some kickbacks when they sell NT servers? Then they'd have to cite as many reasons as possible to prefer it over Linux.
Check out this thread on Apache usage in the pr0n industry (the logic being that pr0n sites take way more hits than "normal" sites), and try this article on Yahoo's experience on traffic/load balancing on BSD. Oh yeah, there's this article on why the pr0n industry uses BSD and not M$. "Welcome to Earth - A subsidary of Microsoft"
This is my opinion. To make sure you don't steal it, it's covered by the DMCA.
Dynamic HTML generation is going to be moved off of the server and down to the client via more effective client-side scripting environments. This will reduce server hits and bandwidth use by about a factor of 10. As this trend progresses, servers will be serving up basically two things:
1) Static pages (containing sophisticated client side scripts).
2) Database rows.
If Apache doesn't adapt itself to this trend, the primary advantages it enjoys over IIS will disappear and with it, Linux as a serious alternative in the entire computer industry.
Seastead this.
First off, those numbers sound way off, but a lot of other people have already talked about that.
:-(, but thats what youg et when you use spare parts :-) ).
IIS is faster then Apache on really big hardware according to what I've seen (and what your doing of course), but the exact opposite is true when you get to more ordinary machines.
Case in point, here in my work I have the Intranet webserver, which is a Celeron 333a w/192mb ram and a Seagate 8gb U2W SCSI drive. It runs NT and IIS4. (Its a small company and I had to spend the budget on more important things like network infrastructure and a backup system)
My testbed machine is a Pentium 200MMX, 64mb, 6gb ide drive (not even UDMA
The IIS machine has serious problems when things get busy. On dynamic pages using ASP, the machine easily gets crippled when doing things like large XML parsing, or anything involving a database (because the database server is not a seperate machine).
The little linux server can do the same things (when ported to php) without breaking a sweat.
Its hardly scientific, but with the applications I designed for the company, Apache/PHP is a far better solution from the point of speed. A lot of this is because of issues with the size of IIS, you can't easily trim it and NT down in the same way you can Linux, so you wind up with a lot of overhead.
Its not noticable on a big machine, but on everyday hardware the overhead does a real number on what IIS can do.
The general concensus is right though, you should see if you can get Dell to send you some demo boxes and do the comparison yourself, because in the end performance will depend on your applications more then anything else.
Just as a parting thing... most of the benchmarks out there that show that IIS is faster are all on big hardware. I would say that how well they do in a quad CPU setup doesn't matter when your only looking at a single processor server for what your buying.
-- "So they told me that using the download page to download something was not something they anticipated." - Bill Gates
I won't bother posting the NT results on this machine since I'd hate to lose the uptime {No, wait, it's actually because this machine won't load NT}: /usr/local/apache/bin/ab -n 1000 -c 20 http://localhost/cypoolmain.jpg
/cypoolmain.jpg
/proc/cpuinfo
./ it, if you'd like. I gotta power it down next Sunday for a move to Phoenix, anyway.
yuma:/home/www/htdocs$
Server Software: Apache/1.3.6
Document Path:
Document Length: 15767 bytes
Time taken for tests: 44.384 seconds
Complete requests: 1000
Failed requests: 0
Total transferred: 16081098 bytes
HTML transferred: 15798534 bytes
Requests per second: 22.53
Transfer rate: 362.32 kb/s received
Connnection Times (ms)
min avg max
Connect: 2 58 427
Processing: 266 796 1327
Total: 268 854 1754
yuma:/home/www/htdocs$ w
8:53pm up 140 days, 20:38, 1 user, load average: 1.63, 0.83, 0.47
USER TTY FROM LOGIN@ IDLE JCPU PCPU WHAT
vanboers pts/2 sedona 8:50pm 0.00s 1.36s 0.39s w
yuma:/home/www/htdocs$ cat
cpu family : 4
cpuid level : -1
wp : yes
bogomips : 16.44
Whoops!!! Was that my little 486/33Mhz thinkpad? Dammit! I did the wrong machine. But, seriusly, do the math: 22.53 * 86400 = 1946592. That's close to 2 MM hits/day for a standard 15K static file.
Feel free to
Cheers.
Linux rocks!!! www.dedserius.com
www.dedserius.com
VB != VisualBasic
I've been playing with AOLServer recently and would like to know how much of a performance advantage it currently has over Apache. (Used to be 10 times faster than Apache)
For those who don't know, AOLServer, formerly NaviServer, was built in 1994 by Jim Davidson and Doug McKee. It was then bought by AOL and GPLed a little later. More info here.
Regret for the past, Is a waste of spirit.
If all they can manage out of apache is 20k-100k hits a day on a modern computer they obviously did something wrong with one of:
I haven't seen a lot of speculation as to the why of those numbers, so I'm going to...
You sell computers. Would you rather sell Mondo Big Ass Computer System with Linux, $10,000 or Mondo Big Ass Computer System with NT/IIS, $15,000? Which ones earns you a bigger commisison/profit? So are you going to say the Linux (BSD, BeOS, whatever) box performs as well or better?
I'm trying to teach myself to set people on fire with my mind... Is it hot in here?
Let me add a little... a fairly balanced comparison of IIS and Apache is here
Keep in mind too, that Apache is not the only *nix web server out there, by any means. It's stable and well accepted, but it's not even trying to be the performance-above-all-else champ. Especially if you need a lot of dynamic content, you might want to look at AOL Server for instance.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Friends don't let friends enable ecmascript.
I independantly tested apache with a threaded-"bomb" -- it mass-requested pages as quickly as possible from 16 IP's assigned to a machine to both IIS in a standard out of box configuration running a ISAPI module, and apache's stock configuration (with mod_so) running a DSO. Both modules did nothing but print off some basic html when accessed, and I can tell you this much:
Both were tested to cycle through 1,000,000 hits. The apache server lost 3 requests (they hung and timed out after 20 seconds), the IIS machine lost 316 requests (same timeout). The apache server was done with 999,997 hits in 48 seconds, the IIS server took 82 seconds to do the same.
Apache won. =)
Anyhow, your results may vary. Both machines were pIII 500's with 256mb of ram, one running Windows 2000 Server and the other running on a RedHat 6.2 install.
When the Mindcraft benchmarks came out, I found this article: What the tests prove. Basically says benchmarks are pointless cause the bottleneck is your connection.
From the end of the article: "Let's be clear about this: if you have only 5 T1 lines or less, a single CPU Linux box with 256 MB RAM will wait on your internet connection and not be able to serve up to its full potential. Let me reemphasize this: ZD's tests prove that a single CPU Linux box with 256 MB RAM running apache will run faster than your internet connection!. Put another way, if your site runs on 5 T1 lines or less, a single CPU Linux box with 256 MB RAM will more than fulfill your needs with CPU cycles left over."
> All in all, I thought Linus had to eat crow
> because IIS really *did* outperform Apache on
> Linux. Wasn't there a story about that on
> Slashdot about a month ago?
Yes it may have but I have to ask...was the main bottleneck found and published?
Remember...we have here IIS on NT vs Apache on Linux. Not only 2 different web servers, but 2 different OSs. There is so much room for different configs here that its not even funny!
Now... Apache on NT vs Apache on Linux - that would be interesting. Even then, there would be the questions of performance tuning - which filesystems were running, what hardware (perhaps the disk controller or some other feature is particularly badly supported under linux and a different one would work better? Perhaps there was some other OS stupidity...like every service known to man turned on)
All in all, its important to know where the main bottleneck is. Is it a kernel bottleneck? some other resource? Is it apache? What modules were loaded in apache? What features turned on or off in IIS? Which compilers were used on the kernel and apache? Was it stock "compiled for i386" binaries of both, or were they compiled all for the specific architecture that it was on?
There are so many variables that could have impacts, it isn't even funny. Im not saying that NT/IIS is not faster than linux...I am just saying that there is alot of room for variation and that all of these things really need to be investigated (on both sides - I just don't know enough about NT to comment on performance variables in it - the idea of a GUI running on my server console gives me nausea though)
The rest - I agree. If you don't have, or don't plan to have the bandwidth (or the interest) to be getting 1 million hits a day, then it really wont matter anyway.
Besides...if you rgetting that many hits, then surely you want to cluster the machines for redundancy and load balancing anyway.
-Steve
"I opened my eyes, and everything went dark again"
-2147467259 means that COM is fucked. (more or less). I'm a Microsoft developer and don't really trust COM that much myself. I've seen really weird things happend to boxes that get to many CreateObjects().
There is a slight chance (in hell) that things could be better on a Win2k box, conpared to NT4, Win2k is a hell of a lot tighter. It's also COM+ based, vs. COM/DCOM, ya may want to check that out.
it's NT4 right?
this is my sig.
I have seen many different benchmarks. Some get Linux out ahead, some IIS.. Tux has been shown to kick butt as well.
The problem with the Apache 1.x series, is that it's not threaded. That means that memory usage under heavy load is outrageous compared to a threaded solution. So - I'm guessing this might be part of the problem.
Apache 2.x will be threaded. You can download their alphas if you're really curios.
Tux 1.0 at Red Hat is a good performer as well. It's kernel based, threaded and extremely high performance. You might want to check that out! Source is available, though the licensing is not listing.. Anybody know?Furthermore, if you get the 2.4 kernel, it also has a kernel space http stack that's used for static content. That way, apache can do what it does best, while all images and static pages will be handled by the kernel space server. Neat, huh?
Stop the brainwash
For example: http://www.micro sof t.com/NTServer/nts/exec/compares/NtLinux.asp
I used up all my sick days, so I'm calling in dead.
Whoops, you're totally right, with regard to serving http. The BSD's were at the top for most of the tests, but Linux was the http king, and that's what the original question was, wasn't it :)
Don't go there
.oO0Oo.
There are places where the networks are not touching,and there are places where they are-Boeing's Lori Gunter
You mentioned in the second Linux Today article that you intend to integrate TUX with Apache. However, Apache has always been a cross-platform server and is heavily used on *BSD and Solaris. Do you feel that this integration will undermine the portability work of the Apache team, or will it simply provide an incentive for web servers to be running Linux? [...]
Ingo: TUX is a kernel subsystem with a small amount of user-space glue code to make it easier to use the TUX system-call. I believe that integrating kernel-based HTTP protocol stacks into Apache makes sense - i dont think this will 'undermine' anything, to the contrary, it will enable similar solutions on other OSs as well.
Much like Linux is ONLY secure if you CONFIGURE and MAINTAIN it properly, IIS is quite stable if you CONFIGURE and MAINTAIN it properly.
Vintage computer games and RPG books available. Email me if you're interested.
Vintage computer games and RPG books available. Email me if you're interested.
get two windows boxen, put apache on one. Slaughter them with whatever stress test tools you have. Check to see which one starts refusing connections and when.
Btw, what do you need these servers for? Does the nature of their proposed application suggest which db you should go with? EG, if your companies data is all stored in MS SQL, you'd be an idiot to go with apache on linux, whereas if everything's on a big fat unix db, why bother fucking with IIS, right?
--
Peace,
Lord Omlette
ICQ# 77863057
[o]_O
The license fee matters. For the price of an MS-IIS license, you can get another box running Apache HTTP Server. So unless well-tuned Apache is twice as slow as IIS, Apache has the lead here.
<O
( \
XPlay Tetris On Drugs!
Will I retire or break 10K?
<O
( \
XPlay Tetris On Drugs!
Will I retire or break 10K?
The current WebSpec99 numbers for Linux and Tux blows away all other competitors.
Look at: http://www.spec.org/osg/web99/r esu lts/res2000q2/
Show your boss that Linux has twice the performance of Windows.
Tux is a kernel module that integrates the html protocol directly into the kernel. It can serve static as well as dynamic web pages and can be integrated with Apache.
Check it out at: http://www.redhat.com/tux/
-- Never make a general statement.
The point here is that Dell made a deal with China. Dell gets richer, the Chineese fatcats on the other end of the deal get richer, and the coolies who work 18 hours a day in Dell's chineese sweatshop get screwed. That is evil.
Now, if the Chineese government desides to standardize it's systems on a piece of free software (Be it Linux, *BSD, whatever) there is nothing that anyone can do to stop them. The difference is that NOBODY IS PROFITING BECAUSE OF IT. Linus dosn't get one cent if the chineese government desides to give each and every one of the ~ 1 billion Chineese subjects (can't really call them citizens) a copy of Linux.
It would be a different situation if, for example, Red Hat were to cut a deal to become the Official Linux Distribution of Red China and had all it's CD's pressed in Chineese sweatshops. You wouldn't advocate boycotting the GIMP because you learned that the Chineese government used it to make all those giant Chairman Mao posters, would you?
Why is it that the proponents of "one nation under God" are so eager to get rid of "liberty and justice for all"?
I think IIS is faster, too. Esp. static content.
However, i refuse to use IIS on my web servers. I use Apache + mod_perl + SSL on linux. My choice has more to it than performance reasons. Apache has better security, modularity, and better development support. I need to use my time for development, not rebooting my servers and waking up in the middle of the night wondering if my server was hacked.
__________
Also, if you want to see a good MT Unix-based webserver, look at Aolserver. Biggest complaint on it is no mod_perl equivalent, but has all the threading issues worked out that you seem interested in.
Oh, well, as long as we have anecdotal evidence, that's all the proof we need!
More proof that Slashdot is not the place to go for objective information (but I try).
--
Sometimes it's best to just let stupid people be stupid.
Besides just being a nice OS, it had just made itself even better suited as an Internet server (for most services). The recent (end of July -- part of the 4.1.1 release) addition of the socket filtering, coupled with the dataready and httpready sample filters should help ANY web-server -- they ensure, the server is not bothered (from accept(2)) until the data-packet has arrived (for dataready) or until http-headers have arrived (more specialized httpready).
The next Apache release will have the FreeBSD code to use this filters (already committed in CVS-tree, AFAIK) or you can patch any other server to use them -- including an earlier release of Apache. Here is an example -- a patch for Apache (and Apache's documentation).
Look for the SO_ACCEPTFILTER option in setsockopt(2) and the accept_filter(9) ).
According to my source, using this on busy sites (where keep-alives are already disabled) may bring down the number of Apaches needed 5-10 fold...
-mi
In Soviet Washington the swamp drains you.
Apache has better security, modularity, and better development support.
That should be suffixed by "IMHO". IIS uses the OS model of security, and the NT/2000 security model is extremely comprehensive. As far as modularity I'm not entirely sure what you're referring to however IIS has the modularity of any ISAPI filter, ISAPI module, CGI, and the script host scripting types (i.e. Add in as many scripting types as you want). Map file types to ISAPI filteres, etc. ANYTHING can be done with IIS. Hell even sticking with straight JScript and ASP you can implement an infinite variability of logic in COM objects. Regarding development support I hugely disagree with that. Firstly if using Perl the standard help holds true with some small changes. If using ASP there are a massive number of getting started help sites, and there is the MSDN which is an extraordinary help.
So in other words while I've said nothing about Apache, I have countered what you mentioned about IIS. Apache might be awesome, but IIS isn't chopped liver itself. It's actually quite a fantastic product however each use has to be determined on a case-by-case basis.
Although I really enjoy the colorful comments from many posting here I seriously question the process that many of you follow to obtain and verify information that could be critical to your business. I run Dell servers and laptops with Linux without problems. I also run NT based servers. The machines in question are appliances. After calling Dell they stated that this was a typo on their website. Both types of machines have similar if not identical benchmark numbers. According to the rep the information will be updated VERY SOON !!! Please call before continuing to post trash here. For those of you with a little more open minds ... check the SpecWEB results at http://linuxtoday.com/news_story.php3?ltsn=2000-07 -05-001-04-OP
I shall have to edit a line in your post to be Win2K compatible:
NT... change ip address (no reboot)... install service pack 1 (reboot: 2 minutes)... IIS patch found, installing (terminate IIS, patch, restart IIS)... DONE.
Get with the times, NT4 is dead as far as an NT web server goes. How about a one-step solution:
Buy Windows 2000 Datacenter Server... DONE!
Too expensive, you say? Then you shouldn't even be in the web server business!"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
No I would not take Commander Taco into the board room, but how about Dell? I remember a couple of months back how there entire IT department was sick of MIckeySoft and went behind managements back and installed LINUX and Samba (unfortunately nothing about Apache). The management congratulated them on how they solved all the reliability problems that were costing them so much money and asked them how they did it. Because of that Dell now sells Linux corporate solutions. Or how about the article a month ago explaining how most of Mickeysofts money making services do not, and indeed cannot run on NT, but how they run on UNIX systems? That might be interesting. A comparision of the latest kernel and Apache 2.x based system will also yield different conclusions, as has already been pointed out. But what about free BSD and AOL server, which is free last I heard? It keeps AOL up and I know that Yahoo runs, or used to run it. If it handles that much traffic reliabily then for sure it is a better bet than NewTerd- Idiot Misinformation Server.
Depending on your architecture's cache design, this isn't always true. Virtual caches that include PIDs in cache tags obviously don't have to be flushed; all that's needed is a change to the cache's PID register. The same applies to TLBs. The Linux Cache Flush Architecture.
Additionally, unless you put in enough memory, you're going to have to swap to disk throughout the rotation.
This is why web servers have lots of memory. But, that aside, in Linux text pages are shared; you don't have 700 copies of your program's code in memory. Clearly stacks and data pages aren't sharable, but your MT system also needs separate memory for each thread's stack and thread-local data. If I had large amounts of data that was usefully shared by my threads, I would shmem that same data in a process-based system.
But the most telling observation is how similar processes are to threads from the Linux kernel's perspective (which you could use to dog on Linux's threading support, but I wouldn't). Both are created by the kernel's clone call, which takes flags that indicate which process stuff (FDs, data, etc.) should be duplicated and which should be shared by the cloned process/thread. If you ever wondered why threads have their own PIDs in Linux...
Threads aren't god's gift to programmers. But they're still nice :)
Get the servers.
Install them.
Send all image requests to anonymous FTP.
"I told you so."
Install Linux/Apache.
Get a raise.
however i was reading that someone (redhat i think) was working on something they called "Tux", that blows the doors off of IIS. I'm not sure when it's supposed to be released but apperently it will be incorporated into Apache.
Anyone have more info on this mystical "tux"??
/* oops I accidentally made a comment, sorry */
Oh, yes. I've dealt with their tech support for Linux. It took three days to get that stupid server up.
Funny thing was, I could have done it myself in two hours if the documentation had been correct. Their tech support was only helpful because the had the right information. Other than that, they were pretty clueless.
No, they don't really care. Then again, what entity is there that exists solely to make money for people who only want to make money that does care?
I say, if you can't get people that care, at least get people that know. Go with VA Linux or Netfinity. That's what I'm doing next time.
I got my Linux laptop at System76.
i have found similar results. the scheduler seems to have gotten a nice kick, as well. under heavy concurrency load, i saw performance increases in the order of 800-1200% (!) on identical hardware.
/me thinks apache + tux on that dell hw will perform nicely...
ydnar
The Cure of the ills of Democracy is more Democracy.
Erlang Developer and podcaster
I was testing on a Compaq Proliant Dual Xeon 800 1 gig of Ram, 8 Gigabit ethernet cards running traffic to 8 separate networks to 16 clients.
With 2.2.X the driver I was testing could only muster a total thruput of around 370 MBPS.
With 2.4.0-test8 ( ia32 )
My Cisco switch was the bottleneck.
Bitcoin pyramid: Join here: http://www.bitcoinpyramid.com/r/1427 it's FREE!
I don't want to accuse Dell of malfeasance, but if they really benchmarked these numbers, they must have badly misconfigured Apache. My guess is that they had Apache doing reverse DNS lookups on every query, and the queries all had to time out because there was no name server responding to them. That's the only way I can imagine that you could get such bad numbers.
> 'Three years without a remote hole in the default install!'
DOS attacks are not root hacks. OpenBSD has never claimed to be immune to DOS attacks. And unless you can control the connection end-to-end (or at least a hop away from your box), you can't ever be.
I've finally had it: until slashdot gets article moderation, I am not coming back.
You could also sell them on security and reliability, areas in which IIS simply falls on its face. A look through old Slashdot stories will give you ammunition. --Tom
Tom Geller
I have two links for you, from an os-agnostic source, the german computer magazine c't:
The first is in only german and show iis and apache are at the same speed (with 2.4pre-kernel IIRC).
The second article is really interesting (and in english). They measured downtime of business class webservers and compared nt to solaris and linux.
To quote the beginning of the article:
To Be Up or Not To Be Up
Analysis of Web Server Downtimes
Stability is one of the major criteria for web server performance. Although it is commonly accepted that Windows NT and IIS cannot match Unix and Apache servers in this field, there are hardly any tests to confirm this assumption. An availability test of the major German internet businesses clarifies the situation.
If you are looking for static page serving, then apache is not the speediest, it's fast, but not as fast as the competition. This is even the cliam of the apache group
For Dynamic serving, the story is a bit different, with apache puling up on par or ahead with the competiiton ( IIS, Netscape E, Zeus ).
if you want the absolute performance king, then the no question winner is zeus, it has been for a while the fastets serving static content, and a great contender on dynamic. The price is not a consideration, since it's not only well worth it, but when you are buying high end servers, a very little add on
Couple zeus with FreeBSD or solaris, and you got yourself a mean combination
.. and I don't know how Apache compares to NT, but since you're obviously wanting to use Linux, why not monitor the situation with Tux and khttpd. khttpd is the kernel webserver, and provides a webserver in kernel space, while Tux is a mix between kernel space and user space.
Khttpd is exceptionally fast when delivering static content, while it gives all requests for dynamic content to Apache. And apache being quite fast at dynamic content this works out well.
Tux however handles both dynamic and static content. It is also exceptionally fast at both. Take a look at this slashdot story.
Ah yes! We're back on this subject again!
Let's get started on the same thing for the millionth time.
FIRST! Dell server hardware can be summed up by one word: wierd. Thier PERC Raid controller blows goat nuts.
Second. Most likley this was tested on a great big honking 8 way Xeon with globs of ram. Yes, the 2.2 kernel has "good" SMP support, and yes 2.4 kernel has better support, but Apache 1.3 DOESN'T!!! It simple doesn't have muilt treaded support like ISS, or what ever. Apache 1.3 runs best on many single proccessor machines, clustered some how, sharing content through NFS. If you want to do the Apache vs, ISS on muilt processor machine bit, test with Apache 2.0. I think it was showed once, and it scored 4 times better then ISS.
Sure, one HUGE machine with no fail over is what you might want for your intranet server where only people will complain. However in a co-lo, running a website that runs your company, many small cheap machines, that can quickly be replaced, upgraded, and scale up is what you want.
Think about like this, you can spend 4k on a small one process machine with a ide drive, or 50k on this huge central heater with globs and globs of ram and a huge SCSI raid system. You'll need two of these space heater, that 100k. For 100k how many of these 4k cheap-o machine can you buy? About 25 machines. For about 4k, you can get a single processor 600 mhz and a 20g ide drive, about 1/4 a gig of ram, a nic, and a rack mount able case. If you only get 20 single processor machine, you'll be be able to flood your connection, and have money left over for a 250g RAID 10 system. My number may be off (WAY OFF), but my point is made.
If a 4k machine blows up, so what, you have 19 other machines. If five blow up, oh well, you have 15 left. If this 50k 8 way space heater blows up, you lose haft your site, and you'll be running at haft power for the next few days until the unsite monkey tech gets there.
MarNuke
If your website is crapping-out on you, the problem is likely to have something to do with your website. I.E. Your application was poorly designed, and the author didn't truly understand the platform he was designing for.
I'm not saying that IIS and MS SQL Server are perfect; They're not. But neither are MySQL and Apache, either.
-----------
"You can't shake the Devil's hand and say you're only kidding."
That's assuming that the hit rate is uniform across the whole day. That's an unlikely state of affairs, even for a site that is of worldwide interest and is getting hits from every time zone. There are always slower hours of the day, and there are always spikes in the request rate.
But you're point is well taken if we narrow the time frame. If we assume that all the visits come in an 8 hour time window, then the best performance (at 100k hits/day) is a mean of about 3.5 hits per second. If they all come within 12 hours, then the best mean performance is 2.3 hits per second.
Now, I've gotten Apache servers to serve up hundreeds of pages per second on fast Pentium machines! If you're averaging less then 10 hits per second over a whole day, then you have severely misconfigured your server.
I submit that the Dell figures are completely worthless without more details about the benchmark. Were they serving static pages, CGIs, servlets, PHP scripts, mod_perl scripts, FastCGIs, what? Were backend applications and/or DB queries involved? How did they tune the server? Did they tune it at all? Did they use a caching proxy like Squid? Without this information, the benchmark results are little better than random numbers.
Always keep a sapphire in your mind
Datacenter is only being sold on certified hardware you nitwit, and please explain to me why anyone would want a 8-32 CPU web server?
ostiguy, mcse
Ok, so I'm a few hours late to the party, but maybe one or two people will see this post. First, a preface: I use Linux, I don't particularly like Microsoft products. With that said...
...how am I supposed to convince my company that Apache and open source is a great way to go?
Well, you probably need to show them that Linux will make the company more profitable. As you said, the suits don't care about open source. And they really shouldn't. Personal agendas (e.g. "Microsoft is bad") shouldn't get in the way of what the company is there for: to make money. The same goes for you (the original poster). Just because you like OSS (and you should) doesn't mean that the solution for your company is to use it. There are far too many variables to consider.
Where can I get -credible- data to prove that Apache can outperform IIS?
That's really not a good way to go about it. If you're comparing an Accord and a Camry, you don't want to go looking for information that will prove that the Accord is better. You have to look at it much more objectively. I remember back in '92 when I was looking for a mountain bike. I decided what I wanted before I even started researching, and bought a bike that really wasn't the best for my money. I think the analogy fits this situation somewhat.
Besides all that, as I'm sure a million other people have mentioned, benchmarks aren't very valuable. You've probably heard a quote attributed to Mark Twain: "There are lies, damn lies, and statistics." Look hard enough and you can dig up factual data to back up most anything. But it's rarely the full story.
-- "Complacency is a far more dangerous attitude than outrage." -Naomi Littlebear
I have found that IIS is faster than Apache. However, IIS runs, for the most part, as a single multi-threaded process (inetinfo.exe) running in kernel space. Therefore when something goes wrong -- The entire process goes down, which kills everything. Ever try stopping the web publishing service on an NT box that is having a problem? It usually comes back with a message saying can't respond to control function, and you have to reboot. With Apache everything is running in user space as multiple processes - Speed Hit Here? - However, it is much more stable and reliable because if one process fails, it can be corrected without taking the entire server down. IMHO - I would take a speed hit as a trade off for uptime and reliability anyday.
The posted question peeves me a bit, since it asks: "Where can I get -credible- data to prove that Apache can outperform IIS?" rather than: "Where can I find independant validation of these results and/or data to refute them?". If you're really concerned about finding the best solution to your problem (web serving) you should find facts and then choose a solution, not pick a solution and then find facts to back it up. That's not fair to your employer and it's not the way shops should be run.
-User
Emacs is for experts. Pico is for beginners. VI is a disease.
Don't forget, IIS runs in kernal space, apache in user space. That gives IIS a boost, but also causes it to take the OS when it dies.
Vintage computer games and RPG books available. Email me if you're interested.
Say you run a superbowl ad with your URL. Say 20,000,000 decide to check out your site after the game. Those hits arn't going to be streched out over the 24-hour period that comprises super bowl sunday -- you'll get hammered HARD at half-time and after the game. If you are running an online brokerage, 99% of your hits are going to come during market hours, with probably 1/2 or more of those hits coming in during the first hour after the market opens. You've got to be able to handle PEAK loads, not just the sustained average loads.
Why is it that the proponents of "one nation under God" are so eager to get rid of "liberty and justice for all"?
Step 2: Show your boss OpenBSD's listing at Bugtraq (read: "...This was fixed 5 months ago..."
Step 3: Use OpenBSD for security (with lowers costs)
Burn Hollywood Burn
If I were you, I'd get Dell (or the reseller/vendor, whoever) to drop off a couple of identical machines, configure one with RH6.2 and one with NT, and let you guys test them. Do not be afraid to tweak the Apache server based on your experience and knowledge. That will show management where their skillsets lie.
Run one for a week or two, move the content, and run the other for a couple of weeks. Then bring up the MS licensing costs, the Apache statistics on web presence, etc, along with an "Oh by the way I'm a Unix guy" and see what they say.
If Dell refuses to supply the test machines, make sure and bring that up to your management- Explain that basically the vendor is unwilling to justify their claim. Then you might be able to pick another vendor such as IBM or Compaq. I understand that Compaq has a fairly liberal test/loan program for such things.
Don't sweat the petty things. But do pet the sweaty things.
This page on Dell's site might also be of interest.
Bingo. Dell itself reports July 25, 2000 SPECweb99 results. Compare items 2 and 7 to find a common platform score comparison.
(rank, vendor, system, score, #CPU's, OS)
2* DELL PowerEdge 6400/700 4200 4 Red Hat Linux 6.2
7* DELL PowerEdge 6400/700 1598 4 Win 2000 Adv Server
Go to the SPEC web site. Then search for the SPECweb99 results Dell submitted.
Apparently, they are using Redhat's Tux server, not Apache. I don't know whether they are related but the combination kicks IIS's ass. You can't get much more definitive than the manufacturer's own tests using the recognised industry benchmark.
This page on Dell's site might also be of interest.
Yeahim sure this is informative, its really good to know, I did not know this, but this seems like a typical response from a Linux Enthusiast.
:) But my first thing to do is to take Linux off and throw FreeBSD on there :)
"Oh oh we dont have this feature yet but it WILL be in next version....."
It just seems like well I dont know, and I know Linux is progressing much faster than it has ever before, but I could go do the same thing with FreeBSD and im sure I can max out a switch.....
The main thing I want to say is just that it doesnt seem fair or rather helpful to this guy that a kernel that isnt even production level does what he needs it to.
I suppose you got modded up because you were showing facts about Linux, but Apache runs on many platforms not just Linux
So I dont really know what the point of the post was, other than the fact that a Dell comes with Linux 6.2 => or is that RedHat 6.2?
Jeremy
It's true, IIS has run in ring 0 since 4.0. I think they did this to get a leg-up on apache/zeus/netscape/etc...
:-P
It provides a performance boost, but putting something complex like a webserver (or browser for that matter) into kernelspace is just asking for it.
Having it take the OS down with it on a crash is the best thing you can hope for... what if it doesn't take the kernel down on a crash, but decides to trample all over memory and data instead? It could misbehave and mangle SQL queries from there on in and it could do signifigant damage before it's noticed. I hope that database wasn't too important...
---
Where can the word be found, where can the word resound? Not here, there is not enough silence.
"Where shall the word be found, where will the word resound? Not here, there is not enough silence." -T.S. Eliot
c't (german magazine) compared apache vs. IIS in a quite exhaustive test and made some very interesting points. Though not Win2k was involved the article may give you some hints and it is even available in plain english here
I also found this bit interesting, if you look at the machine that is configured at the top of the ad it has 256 MB and is quoted at 20K-100K hits, and directly below that is the exact same machine with only 64 MB and surprisingly it is quoted at 20K-100K hits. So apprently quadrupling the amount of available hard memory does nothing for the performance of a machine....uh....er something.
Funny and I thought Perl == Paid employment recently located
I test nic cards and no mater what I use gigabit or fast ethernet. I can't get a total thruput of more than 370 MBPS.
I tried similar test with kernel 2.4.0-test8 and nearly got fired for shouting.
I was able to max out my switch and there is no upper limit in sight with the Cisco hardware I curently have available to me.
The 2.2.X TCP stack is NOT multi-threaded.
The same bechmark on the 2.4.X kernel will take your breath away.
Bitcoin pyramid: Join here: http://www.bitcoinpyramid.com/r/1427 it's FREE!
There have been a lot of good posts on this subject, but all that I've read missed the one obvious problem with Dell's claim (or the post): ;)
100k hits per day is just over ONE PER SECOND. A 486 could do better than that.
Even if you assume that peak time on the site is a 6 hour period, you're getting close to four and a half hits a second, which is no big deal.
For some good information on the state and progression of Linux, look at http://www.kegel.com/mindcraft_redux.html
I know, love Open Source, hate MS. But are you open to the possibility that Apache may not perform as well?
This is not a troll, but a serious point for all--make certain that your loves, hatreds, and desires don't over-rule what may be a valid goal.
If the goal for your project (as specified by the powers-that-be) is performance over ethical ideals, then be certain you know what you're really looking for before you go looking.
But then you know that since you are a techie.
:-(
One of the hardest jobs a techie has to do is convince a clueless boss there is more to a job than a rigged benchmark on a 4-color marketing sheet.
Here are some points to bring up:
Since there are more Apache installations than IIS installations, there are more engineers on the market who understand the technology. You can even throw in a few MCSE horror stories for good measure.
IIS will tie you into an NT platform, and in 5 months (when M$ gets the appellate court to delay their case for a few years) the licensing fees are going to shoot way up. Every major consulting firm has given exactly the same prediction to their largest clients, when M$ wins or delays its antitrust case, licensing fees will increase 2x to 10x, and to reserve a major portion of budget for it. If you choose Apache, later you can switch reasonably painlessly to Solaris, linux, or any other system as costs or management changes dictate.
If you are going to serve only static pages, then IIS wins slightly. If you start to generate dynamic content, Apache blows past IIS. Go search the web for some of the other comparisons.
Apache installations are far more stable than IIS, and there is a lot of anecdotal evidence on the web to help you back this up.
When it comes time to add a custom feature to your web servers, an OSS solution like Apache is likely to have it covered, but with IIS you are at the mercy of M$. If a new feature doesn't exist for IIS, you don't stand a chance of convincing M$ to add it. They have a long history of doing only what they want, not what their customers are demanding.
Get creative, or you will be stuck with IIS, and it will be time to find a new job
the AC
Hemos is like...sci-fi fans;he thinks technology is cool, but he hasn't bothered to understand the science it's based on
http://www.flyingbuttmonkeys.com was a 250MHz AMD k6-2 with 32MB of ram and a 4GB IDE hard drive during the past month when it's been repeatedly slashdotted. Twice, it served over 120000 hits an hour, at an average packet rate of more than 1000/sec, without a problem. Lately it got unreliable because of a failing network card, but that's not exactly Apache's fault. 100k hits a day, right. If the slashdotting had been in effect for the whole day, my little Linuxbox would have taken and served 2.88 million hits that day. the load stayed around 0.20. Imaging what a machine with some actual memory and CPU could do!
________________________________________
Napster-to-go says "Fill and refill your compatible MP3 player", which is a lie. It's not MP3. It's WMA with DRM.
I'm SURE Dell has the documentation to confirm that ad. Call and ask to see it. Dell is very good about things like this and you should be able to look it over yourself.
Also, ask to borrow two boxen for evaluation. They'll do it.
With Apache everything is running in user space as multiple processes - Speed Hit Here?
Yup, The apache processes run round robin (far as I know), and that means that you have to flush the CPU cache for each new connection (rotating the memory space). Note, this won't be noticable for anything but the heaviest loads (since regularly scheduled OS context-switches will have the same effect). Additionally, unless you put in enough memory, you're going to have to swap to disk throughout the rotation. This is likely when you have large mod-perl code (I've gotten apache up to 20 meg / worker-process). Additionally, AFAIK, all IO caching is redundant with MP. I THINK there was a shared memory segment that resolved this, but I don't know.
An MT version has no such problem, since the cache lines are valid for all threads, AND if you had 40 threads with 4 - 12 Meg each, it's unlikely that you'll push yourself out to disk.
We're assuming, of course that nobody is using flat-out CGIs. If anyone was, then Apache would win out-right. In a single-tasked, MT environment, you have to initiate an entirely new process space, and before that, you'd have to configure your environment. In Apache, you have ready-to-go worker-processes than can easily exec the CGI, a new worker thread is recreated at a later time. In Windows, I believe there is less process-creation overhead than in UNIX (assumption based on the non-forked model of windows), so the distinction may not be as great.
I definately agree with the added stability inherent with discrete processes. However, it can actually be harder to debug when you have faulty embedded applications (mod-perl), since each worker will have a seperate work-space, you get a different answer every web-page access. Course, on the other hand, MT perl is even less secure.
With MT, you have all the fun of race-conditions and hidden shared variables that clobber each other. Debugging that can be a nightmare. IIS gives you their equivalant of perl / visual basic. I don't really know much about it, but I speculate that it's more integral to IIS than perl is to Apache, and thus is more efficient. Not to mention, perl wasn't designed for the web whereas ASP was.
Now I know we're not focusing on perl here; you could very easily substitute php, or SSI static pages. I discount servlets, since most of your processing time supposedly is taken up in the servlet, and that's independant of the web-server. I assume that servlets are available (in one form or another) for IIS. Hell, if you really liked servlets, you would use it AS the web server.
I've heard that Apache is comming out with an MT version, BUT, here's the problem. In a windows platform, you still have all the negatives of being in server windows platform. Not to mention, Apache is a text-file configuration while windows is a gui based configuration. The two do not mix well. In the UNIX environment, however, our kernel MT model leaves much to be desired. I vaguely recall reports that showed how poorly Linux did in MT compared to other platforms (obviously Solaris, but I think we even lost to Windows). If you compare MT Apache on Linux to Windows, you might get a nasty surprise. I don't know if the situation has been improved for the 2.4 kernel. Hell, Linux isn't even POSIX compliant with MT.
For simple static-page serving, MT apache should accel. Highly optimized C code in tight loops with a fixed number of worker threads could do magic. This avoids all that BS VC++ MFC stuff. But unfortunately, you can't make a sophisticated business around static pages (unless you're into pr0n I guess).
-Michael
-Michael
Is there such a thing? Really?
As well, what's the difference between "supports" and "speed?" Not to mention cost. (I'm sorry, but I just put two Linux servers in place yesterday simply because Microsoft wanted over $5,000.00 for the software licenses alone; Secure Apache came with Red Hat for $170.00.)
The server I have up seems to foot the bill quickly and correctly. What kind of bandwidth are you going to have? Can your BANDWIDTH support that many hits? Do you expect that many hits? Contrast this again with the advantages of Open Source as you see them.
All in all, I thought Linus had to eat crow because IIS really *did* outperform Apache on Linux. Wasn't there a story about that on Slashdot about a month ago?
-- Talonius
My reality check bounced.
I have no idea what the hell you are talking about NT crashing once a day for an IIS installation, most of our production servers have 4 to 5 month uptimes. I've never seen an NT server that is simply serving pages, dynamic or static, crash. Ever. I have seen the box slowed down, but not because of microsoft, but because of Sun's shoddy JVM and Allaire's crappy JRUN installations.
I've witnessed and worked on deployments of sites that bang massive loads, and you know what? If your site is pulling a million hits a day and you're attempting to box that into one server, then you're a complete moron and deserve to suffer those crashes. No OS is going to save you from that.
For huge sites, the network design is just as important as the application design, and if you fail on implenting a solid design - regardless of what OS you are serving off of - you're going to run into problems.
This was taken from Apache's own Performance Notes Site.
IIS may be faster. I actually don't know, because I've never used it. But I will say this: I worked at LinkExchange when we were the number one company in Internet reach (52%). That's right, more eyeballs than AOL, Yahoo, and MSN combined. And we did this using Apache; both for our site, and for our banner network.
Not to start a flame war, but there's also a chance that the performance bottleneck is Linux, and not Apache - LE was using FreeBSD. There's an excellent benchmark of various Unices, which may indicate as much. It well done, but doesn't get going until page 8 or so...
Anyways, be sure to take administrative costs and bandwith constraints before making a decision.
This shows how a bleeding edge webserver, TUX 1.0, running on a tweaked 2.4.0-testX box can outperform a virtually identical box running IIS 5.0. Curiously enough, these are DELL boxes, and the tests were performed by DELL.
I understand that it is hoped that the advanced features of TUX 1.0 will eventually make their way into Apache.
Apache has been designed for correctness and it *chugs* compared to IIS, Netscape/iPlanet or Zeus. But it's stable and works well.
If you need a speed boost try putting a Squid proxy in front of it - it'll really help on static pages / images.
You may find that the Apache 2 Alpha/Beta software performs better than the 1.x line.
If your web server is down then it ain't going to be handling any traffic now, is it?
Do you *really* believe the numbers and statistics handed to you on the glossies? Wouldn't it make you a rather credulous person if you did? These are *salespeople* that make these things up yyou know.
Government of the people, by corporate executives, for corporate profits.
Where can I get -credible- data to prove that Apache can outperform IIS?
I know I'm risking a troll rating here, but shouldn't the question be "Where can I get -credible- data to prove which is better?" If you don't have access to the information that Apache kills IIS, how do you know that it is, in fact, better? I'm not saying you should take Dell's advertising at face value, or that IIS is better, but you are presupposing the answer to the question that you (admittedly) don't have the answer to. I suspect Apache would be better on many, if not all fronts, but I don't have any data to back it up either.
+5:offtopic,but anti-American