Slashdot Mirror
NT Beats Linux in Round 2
strat writes "PC Week ran it's own benchmarks, with Mindcraft, Microsoft and
Red Hat. The margins were tighter this time, but NT still fared
better. They specifically mentioned the lack of multithreading
in the Linux IP stack as the main bottleneck. I wonder how
4.4lite would have fared? "
What is 4.4lite?
Oh yeah, 1st I think.
The neat thing is the Linux community's ability to nimbly respond to things like this with actual physical changes, whereas Microsoft's respnonses are innuendo and smokescreen with real changes maybe coming much later.
With all due respect to BSD (I am confident that BSD would have kicked MS butt) ... the test was not free software vs. Microsoft, or ABM (anything-but-Microsoft) vs. Microsoft. The test was a specific replay of Linux vs. Microsoft.
The problems have already been identified and will be fixed in 2.4, which Linus is hoping to release in fall.
Linux was tested on Microsoft's home turf and lost. We penguins can whine about the unfairness (I am still inclined to believe it was an unfair test) or we can roll up our sleves and fix the problems (I do believe that the problems are real).
Let's work to get better IP and better SMP scaling for 2.4.
Ok, what have we learned?
1. Don't fight MS on their terms.
2. There's a couple places in the kernel that need touched up.
Can we get on with our lives again or will we dwell on this forever?
I would argue that a 4 CPU machine with 4 net cards serving only static content is not realistic at all.
It's just a thru-put test of a semi-fine-grained-SMP-kernel.
I wonder how Linux and NT would have performed using a gigabit ethernet...
-T
Linux performance has always been near the bottom of the unix world. Even in terms of free unix, FreeBSD smokes it in terms of web performance.
Linux is more apporachable than FreeBSD and hence is more attractive to new users.
It is a loss to the Linux community that they take much interest in the NT vs. Linux performance test. The argument that Linux is only a side effect of the hatred against Microsoft is strengthened. I hope that the BSD community takes no interest in such performance test and that they value their os as it is.
How is *BSD's SMP support?
No.
It is a loss to the Linux community . . . yada yada yada.
If you read any of the above posts you'd see that the common attitude is Linux lost, why, and how can Linux be improved.
It is helping us to improve Linux.
Why do you BSD posters commonly act holier than thou? I'd actually like to try BSD someday despite the whining of several BSD posters who seem to have a chip on their shoulder about Linux being in the limelight for the present moment.
Linux lost the battle. It was tested on MS turf and, surprise, surprise, it lost.
:-) Well done. Now if Inuit and IMSI would just wake up and port to Linux I could dump MS products for all but the occassional game.
There has been a lot of banter about why it lost, justfiying the "unfairness" of the tests, etc. I don't believe the tests were accurate, but still...Linux lost. We can count on the people who write this wonderful OS to fix these shortcomings fairly quickly. In the meantime, would those people who are quickly finding reasons to reject the tests work just as hard if Linux had won? They likely would have beat on MS for trying to put the results "in perspective". It's happened before.
Still impressive that a "rag-tag" group of hackers and crackers can write an OS that only loses by a little on MS turf
Yes.
Its time to show em how the Open Source community works. The problems have been identified. Now fix 'em fast and make it better than it was before!
Could you post it a third time tomorow for all of us who missed it the first two times and for those who wish to dwell on this forever.
Just remember, Microsoft is a huge corporation who can afford the resources to do everything they can to win certain benchmark tests. It's not like the Linux community is out there coding just to beat Microsoft in tests, but to make an overall better OS overall. It takes a little time, but in the end we'll have the ability to smoke them - and then some.
He posted a link to the previous time it was posted... You can't argue with that.
Instead of whining "Cock sucking the geeks in Redmond" and other unprofessional, immature such tripe:
Come to realize:
1) That benchmarks (like a college degree or certification) are a 'right of passage' or a check off item for people who are unwilling or unable to test the machines in their own real-world application.
2) Companies know #1 and work to make thier code run the best on the benchmark in question.
*IF* GNU/Linux wants to play the 'benchmark' game, then GNU/Linux will have to do #2.
If your goal is to be the best damn penguin you want to be, then don't GIVE A DAMN about the benchmark.
Either way, fair or unfair, NT won the benchmark. Its up to CODERS to either do #2 or be the best penguin.
The BSD camp *DID* get breathing room to get their numbers up for the next benchmark test. And for that, they should be thankful. For if they win, they can then brag how they would have won from the get go. I know *I* asked mindcraft in my letter why they didn't test BSD with Linux vs NT.
Funny. My desktop box is a P120 with 32 meg RAM. Works fine for me. Where'd you get your info?
>By breaking up the data volume into four NTFS partitions (each with its own
>transaction log), we were able to distribute transaction entries, eliminating the
>bottleneck
So, NT had a nice set of striped drives to ride on which (duh) sped up HD access vastly, while that loser boy Linux was (unnecessarily) saddled with a single drive filesystem.
So, the RIGGED test PROVES Linux to be the LOSER! But I would expect no less from ZD and its largest advertiser MS.
Call up Hawking, I think we've discovered a new force in the universe that binds the FUD mongers together more strongly than the strong force. As they approach one another they start exchanging Dollarons (a new spin(doctor) 1/2 particle) and become inseperable.
:/
You can't make physical changes to software, software is not a physical thing.
First off I could care less about these tests. If M$ wants to feel good let'em. They will only hurt themselves.
;-)
I work with Linux and WinNT almost everyday. They both perform in speed about equally (give or take on job). But at work, its stability. Lets do some test with WinNT and Linux after a few dozen application installations and uninstallations. Then we will see the real OS.
Microsoft is just a greedy, manipulating piece of shit who cares about nothing but money! Notice how Microsoft tries so hard to put down linux but why? Maybe because they fear losing the market to someone who gives out their software for free. Maybe Bill Gates will take linux and hide it within his binary code like he has done to everyone else and their software. Bill Gates doesn't lie but he does manipulate situtations to his advantage. Ex. 4 network cards. Who gives a shit! Any amatuer could hunt for a couple of days and implement something within a seperate OS that linux can't do or do as well. Why don't they face linux head on with something that matters or for something that it is known for. Instead Microsoft creeps around looking for bullshit. If NT is so good, then Microsoft should do the benchmarks again after linus implements 4 network card optimization. What? Is that not fair to Microsoft? Well let them change theirs too! Oh I forgot, that could take years not days. Can Microsoft update their kernel as frequently and as easily as linux can? Hell no. If they do, their will be a 200 dollar service charge for their previous fuck up! NT may be a dragster, but like dragsters they last for one race and that's it! Linux is like a viper, it can go fast as shit, look pimp, and stay together to drive again tomorrow and the years after. Microsoft will not manipulate me!
pcWeek -> ZD -> M$
PC Week, and really all the Ziff Davis rags are very M$ biased.
Microsoft is one of ( if not THE ) Ziff Davis' biggest advertiser. The rest of their money comes from Widows Related software ads. Do you really think ZD is going to print an M$ stinks article?
Also, I believe M$ owns a BIG chunk of ZD (stock-wise.) So if you're expecting a balanced test and story from an ad-heavy trade rag conglomerate, you might just be a little naive.
Also
>Ok, the results are in. The benchmarks, when
>administered properly, show that NT outperforms
>Linux in every category.
"Every category"? As I recall, there were exactly TWO categories:
1) Web serving static documents using multiple network cards;
2) File serving using Microsoft's proprietary SMB protocol, also using multiple network cards.
In the former case, see what c't magazine found:
http://www.heise.de/ct/english//99/13/ 186-1/
Namely, Linux outperforms NT when a single 100 mbps network card is used. When serving STATIC pages, NT outperforms Linux when using multiple network cards.
Regardless of the number of NICs, Linux greatly outperforms NT when dynamic content is used (CGIs in this case). Almost all websites use dynamic content.
The tests weren't run on file serving, but it seems likely that a Linux networking bottleneck is the culprit there as well, when multiple NICs are used. Other ZDLabs tests show Linux + Samba outperforming NT on single-NIC systems.
Keep in mind that the Samba team had to reverse-engineer the Microsoft protocols, and it's possible MS is doing things specifically to sabotage Samba (they have done so in the past).
I also take issue with your statement:
>The benchmarks, when
>administered properly,
That seems to read:
"The benchmarks, when administered so NT wins, show that NT wins."
There are clearly numerous benchmarks other than Winbench and other than static webpage serving which show Linux comes out on top.
Also remember than Mindcraft has at various times been commissioned to show Windows beating Novell and Windows beating Solaris (yes, Solaris, as in "Oracle will pay you one million dollars if you can show NT+MS SQL Server-7 is no more than 99 times slower than Solaris+Oracle [which we're running on an Ultra 10000, of course]").
>The benchmark has exposed an architectural
>flaw/oversight.
Apparently we need a multithreaded TCP/IP stack. This is mainly of use to people who need to serve more than 100 Mb/second from a single box.
Unless you're serving static content over an OC-3 line, you're not going to hit this snag as a webserver in the real world.
If you DO need to serve a lot of data, a cluster of INEXPENSIVE Linux boxes will do nicely (one CPU and network card each). This is also far more fault-tolerant than a single Quad Xeon overpriced box. (Remember that extra CPUs didn't help out much with static content -- NT with one CPU served 1800 pages per second, while with four, it served 4100 -- not the 7200 you'd expect.)
Four Linux boxes will serve 1300 per second each, for a total of 5200 requests per second.
A few other changes with regard to Apache, in conjunction with kHTTPd, should help Linux exceed NT's static page serving even on single-CPU systems. This isn't terribly important since once you start using dynamic content, you're more likely to see 20 page-views per second per CPU than 1,000 no matter what the operating system.
>We need to fix this in order to reach the
>performance numbers we need to be a server
>operating system.
Uhh, Kostya -- Linux has already reached those numbers in real-world situations. Should it be improved? Yes, and this will happen anyway. In specific benchmark configurations, it's slower than NT -- but if you were to rely on this data and choose NT, you'd quickly run into the problem that NT doesn't have the stability you need for a server operating system. It'd also be much slower once you started using dynamic content in your website.
>So let's do it. Let's fix it. Start downloading
>and start coding.
Somehow I suspect that "Let's" means "Other people". Are you "downloading and coding"?
FYI, people ARE already addressing this issue, but it is nowhere near as pressing as you seem to think.
>I have linux installed on every one of my file
>servers. I will be able to fend off the criticism
>for now.
Criticism from whom? Your pointy-haired boss? Offer to use NT on a system and see how well it does. I don't know ANYONE who prefers NT over UNIX in any real-world setting for servers when experienced with both.
Are you getting insufficient performance from your Linux servers? I know a lot of people who rush out to get the latest and greatest and then don't make use of it (e.g., a P3 500 for Quake won't improve your game so much as some extra practice!).
Chances are, when the multiple-NIC issue is fixed, you won't even notice the difference. It won't affect most people!
How many tests in a row will it take for you Linux crap bigots to finally face the music? You made unsubstantiated claims of being faster and more stable, yet when it is done with the supervision of Linux people, you still cry foul. I see many comments about changing the partitioning to span the NT log over multiple volumes. You fools overlooked, conveniently enough, that the RAID was changed for Linux because the performance on the default RAID was not a good performing driver under Linux.
Hmmm...so we have a system (NT) that runs faster than Linux on BOTH RAID configurations. Seems a major flaw in Linux (and it always has been there) is a lack of substantial hardware support.
You bastards can cry and whine and scream foul, but reality is that your boasting and unproven claims have just been proven false. Wake up and smell the coffee, fools. Linux is a little toys system for A.B.M. techy geeks who are incapable of functioning as professional system managers.
A professional system manager is one who, regardless of person preference, realizes the best system for the right job. Be it Unix, BSD, Linux.......OR NT.
I use Linux and NT everyday and I can say from hard core experiance that our NT servers with nearly double the processors are no where near as fast as the Linux servers.
Ohh well, like I always say, benchmarks prove only one thing; Benchmarks are useless.
BTW, has anyone heard of a series of NT web servers crashing and wiping data? I read on another news group that some big hosting company really got screwed when all its NT server crashed at about the same time. Is this some kind of attack, or just typical NT instability at its best?
Strider Centaur
If you want to know weather NT+IIS or Linux+Apache is a faster HTTP server, than this is indeed a useful benchmark. If you want to know which OS has faster file IO, then this is not a good benchmark. It all depends on what you're looking for.
This test just HAS to be done with FreeBSD. Will probably beat the hell out of Linux :)
The short explanation:
-linux does not have a multithreaded IP stack
What does that mean?
-there is a single process that accepts and processes IP data.
Why is this a problem?
-it is not a problem in machines with single processors or single IP interfaces.
-it is a problem in machines with multiple CPUs and multiple IP interfaces.
In the multiple CPU multiple IP interface case peak performace is obtained by using the multiple CPUs to process IP data in parallel.
This is a small particular weakness in the kernel architecture that these tests exposed (some would say even exploited). This weakness has been noted, and is being corrected in the next kernel release.
Note also that the amount of data required to expose this weaknesses is nothing short of huge and the tests that exposed this weakness were not really limited by any other component of the operating systems.
The fact that the original test results claimed a 400% (or so) NT advantage seem to indicate they knew exactly what they were doing. The results were expected to reflect the IP stack threading properties that they were designed to expose.
We should all thank Microsoft for spending their $ on finding problems with Linux and allowing them to be fixed. Microsoft please feel free to spend more $ on research into this problem.
It does look good for Linux that this seems to be the most exposable weakness they found.
That's exactly what I'd like to know. I consider one of Linux's greatest strengths to be the fact that it can run on limited hardware and still get a job done. OK so it's not much of a job perhaps, but just the fact that anything's getting done at all, and that you can do it on hardware you would have thrown out otherwise (and lost your investment in), ought to say something.
Let's see some of these tests carried out over 12 months. On a 486.
I laugh at some of posters exclaiming "I work at fortune 100 company and look, NT can serve more pages, my boss says do a mass install of NT." Ya like right.
Please correct me if I'm wrong, but I haven't seen anything in either of these tests about security--one of the major criteria for choosing a server OS. How do they stack up against various kinds of crack attempts? Or are they saving that for Round 3?
I've been working on an NT box (pentium-166, 256MB) -- it's useable, but a bit sluggish...
Function of Linux is not only file or web serving to Windows machines. The total consistency, reliability, efficiency and capability is more important. From that large point of view Windows NT is not even comparable to Linux.
Guys, we better keep cool. Linux is free, and it is here since we want it. There is no possibility of bankruptcy for Linux. Linux is *not* competing with Windows. Windows is competing with Linux. The MS guys should be stressed out about this competition not the Linux community.
It has been more than 4 years since I switched to Linux from Windows. I know what I am using. I know what I was using.
Doesn't Microsoft own 20% or so of Ziff Davies?
/Curious George
You should have been so much scared and desperate. But do not relax, the end of Windows is approaching. There is no way to get rid of Linux. Becase we want it, use it and make it better everyday. Like it or not.
Anti-Linux news: ignore and call "FUD".
Pro-Linux crazy rumors: embrace as the gospel according to Linus.
Yawn.
Er...they did the tests in conjunction with Red hat. Red hat acknowledged the short commings of Linux, why can't you?
I'll give this shortcomming a one-week life span...I'll be very interested to see the results of the same benchmarks run after this performance bottle neck is fixed.
When I first read this, I was shocked; everything that I had read and done myself told me that Linux was 'better' that NT. Well, maybe this isn't the case right now. Right now, I say, because all of those people who love this OS are working night and day to improve it. It will only be a matter of time before Linux surpasses NT. It is OK that Linux is not the best, because it will be one day.
Look how far Linux has come in just two years. At this rate, Linux will surpass NT in these fields within the next two. I am excited for the future now that Microsoft has 'strutted its stuff' for us to see. Let the games begin!
hehe, I totally agree, Linux users are such babies, suck it up /. readers Linux lost. That would make you all the LOSERS. Maybe Linux will do better next time, not that it matters cause every knows FreeBSD is the boMB! LOL.
10. FUD
11. FUD
12. FUD
13. FUD
14. FUD
15. FUD!
>Anti-Linux news: ignore and call "FUD".
/. readers (as described above), and yet choose to read said comments and bitch about them anyway.
>
>Pro-Linux crazy rumors: embrace as the gospel according to Linus.
People who already understand the behaviour of
>Yawn.
Indeed.
As I understand it, they found that they needed to have the data spread across 4 partitions or a bottleneck in NTFS limited through put. Fine.
However, doesn't this change affect users? On a Unix system you can move directories about and make links to the new locations; the users don't have to know or care - it's transparent. NT isn't quite flexible as this right? (Or am missing something basic about NT?) When make 4 partitions on NT, now the users have to know that things are on F: G: H: I: etc. In Unix you could spread the directory tree across different filesystems without impacting the users.
It sounds like there is another fine point about that the benchmark that they didn't mention - not only is the data divided amount 4 partitions but it would seem that the requests are to. In other words, for the sake of speed you have to tell the individual clients that they only can see a subset of the data. (It seems that if all the clients were requesting data from the same partition that NTFS's bottleneck would reappear.)
Somepeople here almost sound like religious fanatics. Its time to face reality. Linux, like an OS has flaws. The correct response is not to deny them, but to examine them.
Check the TPC Web site. The OS has to be very stable to complete a TCP run. Oh looks like there are exactly 0 results on Linux. I wonder why?
Read PCWeek's headline again. Does it say "NT soundly trashes Linux yet again"? Or "Linux shown to be total crap"? No, they take this upbeat stance about the testing showed the Linux zealots where the need work. All over zdnet you see this same fawning over Linux.
BSD had Linux beat years ago. Solid code, designed by real OS researchers. Freely available. But the better technology suffered from inferior marketing, and so now Linux gets all the press. This is a terrible trajedy, IMHO.
Actually, the people doing the Linux tuning were not ZD employees. So - if ZD are a buncha incompetent pricks, that plays no role. Let's just thank the Linux community for creating enough pressure to get this chance to closely compare Linux and NT. Hey - now we know more about what is wrong.
I suggest you do a little deeper reading into what a multi-threaded app or kernel actualy /can/ do. Take a web server that uses POSIX threads for instance (POSIX Threads, or pthreads, is available on most commecial OSes, such as Solaris, IRIX, Digital, NT, etc). First off, spawning a thread is a very lightweight operation when compared to fork()'ing a child as most web servers do (including apache). This in turn speeds up the responsivness of the connection for the client, and the OS doesnt expend an inordinate amount of time and resources to fork() the child. This then translates into the OS having more resources available to handle more threads (in Solaris, you'll see them referred to as "lightweight processes").
That's not all. A mult-treaded app (or kernel) can do different work simultaneously on two or more processors. For instance, if the Linux IP stack was MT, it could use one processor to, say, accept a new connection, and use another processor to service another connection (sending, receiving data)... all at the same time. Things are done in a truely parallel nature, and not serialy, one-by-one like it is now.
Actually, I find an interesting paralell between Operating Systems and Car engines. I know that most of these kinds of comparisons are silly and inaccurate, but let me explain.
Back in the early to mid 80's, US car manufacturers saw heavy competition from low cost japanese imports. On top of the competition, they also had heavy regulation from the government for things like polution control and gas mileage.
In order for the car companies to add the things that the government (and consumers) wanted, they ended up bolting everything and the kitchen sink onto the engine rather than redesigning the engines from scratch to perform better.
As proof of this, I suggest you look under the hood of any mid-80's american car. You can't see the engine! There's so much crap in there that it's hard for anyone but a mechanic to comprehend what it all does.
Now, take a look at a modern day car under the hood. The engine is clearly visible with a minimum amount of extra crap hanging off it.
This began happening around 1990-1992. The competition from japan started to back off, and US car companies started to take advantage of new technologies to completely redesign the engines so that they could go 100,000 miles without a tune-up and have next to nothing in terms of polution control.
Now, how does this apply to Operating Systems? Easy. Microsoft has been on an uphill battle for the last 10 years in it's operating systems. Granted, most of their competition was from themselves. First, they had to win the war against Dos and get people to upgrade to a GUI. Then they had to win the war against 16 bit software and move everyone to 32 bit software. Finally, they, needed to get NT established and a viable OS so that they could win the war against Win9x.
Now that NT is well on it's way to replacing Win9x, they've taken the time almost completely redesign the OS (Win2000).
I wouldn't say they sat on their laurels due to lack of competition and are only now doing this because they have competition. I'd say that the current lack of competition (even from themsleves) is giving them breathing room to stop bolting things on the side and actually redesigning the OS so that it doesn't need the things bolted on.
Here's one case (not a networking stack situation but there are analogous things in the network subsystem I suspect..) where multithreading helps even on a single-CPU system: :-) ):
Single threaded IO system (horribly inefficient
-> Process A requests a read() from hard drive 1. The kernel blocks A and begins reading.
-> Process B requests a read() from hard drive 2. B blocks, waiting for the IO mutex to become available.
-> Process A's read completes. A is unblocked and continues.
-> The kernel begins handling B's read.
-> Process B's read completes. B is unblocked and continues.
With a multithreaded IO system, things could proceed as follows:
-> Process A requests a read() from hard drive 1. The kernel blocks A and begins handling the read.
-> Process B requests a read() from hard drive 2. The kernel blocks B and begins handling the read (in a separate kernel-level thread)
-> Process B's read completes. B is unblocked and continues.
-> Process A's read completes. A is unblocked and continues.
Notice that although B's I/O job was significantly faster than A's (for whatever reason), the single-threaded case forced B to sit and wait until A finished. In the multi-threaded case, B could finish immediately.
Daniel
neither will linux, it is ix86 where x>=3
Erm... did you notice that the german c't magazine is _extremely_ Linux biased since the hype came up? Everything MS does sucks, everything about Linux rules. This is getting boring!
Like some other guys here said: BSD is da real OS... Linux is a mess.
Don't believe the hype!
Seems like Linux is some sort of virus that makes MS sneeze once before they become immune and makes 'em get even more power than before!
Linux is nowhere. You're deluding yourself. Linux is not used in most businesses, period. Notice the hoopla when Burlington Coat FActory was considering using Linux for POS boxes? That wouldn't be aroudn unless Linux was really nowhere. It's only on geek boxes at home. It's not used for mission critical applications. Stop fooling yourself. Linux is the HAM Radio of the computer world. A cute hobby, but nothing else.
Get a clue.
If c't "tested" the taste of a juicy steak vs. a pile of shit they'd prefer the shit if this was en vogue at the moment!
Minute difference in stability and performance are not important to me. Nor do not have a 10Gbps pipe to Internet. Like you say, Linux lives in the real world, not in the suppositional worlds. When I choose my software, I choose free software. Why? Because it make sense: you stay free and you are not a slave to Microsoft. How much the software costs is of little consideration to businesses that wish to be free Microsoft.
As an end user, I am willing to sacrifice even more. I will trade a great deal of stability, performance, and other features for freedom. To software I can trust. To software that are recompilable. My real-world benchmark revolves about freedom.
Use only free software. Haven't you heard the revolution has started?
Guess what... NT would rule again! Because they cheated... HA... PENGUIN-CRYBABIES!
i'm surprised I haven't heard anyone suggest this (or maybe I missed the post) but why doesn't anyone run a dollar for dollar test of linux vs. NT. Give each side $3000 for *everything* and see who comes out on top. For the extre cost of NT i'm sure you could balance things out with hardware on the Linux end.
not if they didnt know which was which.
no money. :( linux cant be tested since no one will pay for the tests. would you like to contribute ?
My jaw droped the other day when I saw windows2000 ads here at slashdot when I was reading pcmags benchmarks. Is Rob involved in this as well? I believe I already saw this url and microsoft is trying to control information like zdnet, msnbc, nbc, adn now slashdot. If I were Rob I would refuse to have ms sponsor my site because ms would twist both of my arms in regards to what I could say about linux and NT. I always wondred why BIlly boy would want all the cable companies and now I know why. In future releases of server I can imagine just what I can or can't put on my web page. I believe microsoft is aproaching the stage of an actual government that believes in sensorship and control like China. Chairman Gates and Mao would be great friends. BIll expects to be the first trillionair within the next 3 years. He will make more then any government in the world by then including the United States itself and will undoubtingly will use huge percentage towards false advertisments and benchmarks so our bosses can force agaisnt our will to use NT at work and reward Bill with more money from fraud and lying!
I have about had it! Is slashdot really going down the road of zdnet?
If this is the future of slashdot and the end of linux as we know it then I will hapilly switch careers!
actually BSD (running MacOSX) basically lost too. The only one who held its own was solaris. so much for BSD.
its the competition that helps. most linux advocates dont just sit there -- they ask the kernel dev guys what went wrong. and that's what keeps the pressure on and accelerates linux development. thats also why linux will beat NT on the next round.
I already mentioned this in a previous post but I was reading the previous fud here and I noticed an ad for windows2000 and my jaw droped. I now know why Bill is buying calbe companies and media. Its control. I believe microsoft allready has the power of a full government. BIll expects to make a trillion a year in 2-3 and ms could easily pay the coutnries debt and even pay for funding. Within 2-3 years he could buy the US government or pay for an army thats bigger then ours. I doubt Bill will really do this but the fact that a mere man has the power to do this is frightening. It seems bill loves censhorship and would be great friends with chairmen Mao. I have a distinct feeling that ms is behind all this and will use slashdot to spread more fud.
I think that guy who wrote this analysis is right, but didn't catch what a benchmark is. It doesn't necessarily have to reflect a real business situation, but merely to test a piece of software in somewhat extreme situations. I don't say it's right to do so anyway, but whenever you read results of a benchmark of a graphics card or a hard disk in a magazine, do you really think those digits are even close to your daily usage ? No, of course, and even though those guys at Ziff Davis must have worked hard to find a benchmark where NT would make Linux ridiculously slow, this is part of the game, and instead of weeping, why don't we work and write nice apps for Linux instead ? This is just what Linux lacks now.
You're either a troll or an ignorant pointy haired bafoon. I take it you've never looked behind the scenes at most major Fortune 500 companies. The boxes are already there, most people just won't admit it.
Let's face it: who cares if NT is faster than
...
Linux in a 4 CPU system, with 4 NICs.
I will never ever have a 4 CPU system.
99% of the persons work with single CPU systems
(even in servers) and maybe dual CPU systems.
Linux is way fastar than NT in single CPU systems.
Let's start advertising benchmarks where Linux
kicks NT on single CPU systems.
On the other way, I'm starting to see some
things in the Linux world that I don't like:
Gnome and KDE are becoming as bloated as Windows.
Example: Yesterday, added three applets to the Gnome task bar.
One applet was a clock, the other was a CD player
and the other was a CPU load graphic.
I started to notice that the system was slow.
I used "ps auxl" and found that every one of this
applets was spending 3 to 4 Mb of RAM (RSS)...
Can anyone explain to me why does a simple applet
uses 3 to 4 Mb of RAM ?
I'm been a X/Motif programmer for a long time,
and I just cannot imagine what are these programs doing to spend so much memory !
I've seen entire CAD and spreadsheet programs that
use less memory.
I would expect that each of this applets would
spend 100Kb, 300Kb, at most 500Kb of RAM. But
never: 3 to 4 Mb
If the Linux word continues to evolve like this,
it we become equal to Windows.
>You're not really telling the clients they can >only see a subset of the data, you're just >telling the web server that it has to grab it's >data from four different drive letters.
:-)
I thought that this was important to the SMB performance too. It seems that the client request need to be spread across the 4 partitions - i.e. unless the partitions are just mirrors of each other, then the benchmark's value is limited by the fact that the clients aren't allowed to look at the same partition all at once.
This is just another footnote to the benchmark. People know (or should know) that benchmarks are often limited to very specific cases. NT won, mindcraft's numbers weren't totally bogus, but microsoft did set the rules of the game - they had "home court advangtage".
OH well, I don't use linux because of how it works one a $50k box.
the 16MB/32MB makes a huge difference. NT4 typically loads into about 20 MB, so you just passed the threshold where the OS needs to swap out. NT is unusable with 16 MB, but reasonable on 32.
its so sad, no matter what linux should be faster
,it was always faster than any windows box!, you wont reach that status again, because NT is catching up and as always there is only so far you can reach with good code. You will hit a limit of achievable fast code.
1. default configs should be fast
2. apache has to be threaded, this fork stuff is the 70s way to program, get real!
3. SMP has to be handled better
Remember 1992 ? when you ran linux and no matter what config
What linux needs perhaps is some of those hacker/demo coders to do some serious speed lifting.
You CAN get speed with stability, just code it right.
Why cant redhat get of their ass and do their
own tests?
Face it, MS isnt as lame as you thought, with 1000x programmers/engineers and smart people paid $90000 + shares, you can achieve good results.
eventually MS will make good code, then what will you do ? cry?
get real!
I think, that's you looking for ;-)
Frank
Do you know what computer is, kid?
My simple question is:
Why the tests of Mindcraft and ZDnet that had so little
information for the user were both reproted on Slashdot.
And at the same time the highly informative and much more
comprehensive tests of c't from Germany were ignored ???
As a user that wants to decide what is the best OS for my
application I am interested in benchmarsk of computers
with a single network card and a single CPU. And I am
sure that I reprisent the vast majority of users. The c't
benchmarks give that information. They also give information
about multi-NIC,multi-CPU configurations. So their
benchmarks serve all the users. In my category Linux wins
hands down. Whoever uses different configuration is free to
chose what suits him best. I've been using Linux for over 2 years
and I love it. I knew about its single threaded TCP/IP stack
before I first started with it. I also knew that it would not affect
my configuration. I've chosen it with full knowledge about what
it can/cannot do. Trolling with the poorly conducted tests of
Mindcraft and ZD net does not serve the users.
All in all, Slashdot is becoming too much noise
and too little signal.
Stefan B.
The dumbasses in Redmond want us to think we did!
NOBODY is using 4 NICs, NOBODY! Did you hear me?!
The issue of multitheraded ip stack was never raised on kernel list! Nobody was even thinking of implementing it! NOBODY ever asked about it.
Anybody who's going to work on new stack implementation is plaing MS hand. PLEASE, don'twaste your time! THEY want you to stop doing whatever you do and join the race.
THEY HOOKED YOU, STUPID!!!!!!!! Red Hat didn't know what they are stepping into? STUPIDS, WAKE UP!!!
I see it every day. Linux faster than NT in a singel NIC/single CPU box (not to mention how much more beautiful :). I just don't get what the fuss is all about. I knew about the limitations of the single threaded stack for a loong time. Doesn' bother me at all. It'll get fixed but I think it's a low priority. Multi-NIC, Multi-CPU configurations are not only more expensive - they are also less reliable. 4 boxes with one NIC/CPU each are twice faster than a single quad box and are also (as a whole) cheaper than the quad and failover comes for free... Someone please tell me what the fuss is all about.
puzzled
Stefan B.
a) In the c't test they had a 50000 $ machine (with 2 gigs ram).
...
:-)
If you take such an expensive machine just for webserving,
why not can you put 2 or more ethernet-cards in it?
(These cards are not so expensive.)
b)Don't forget redundancy! If you have 2 (or 4) processors, 2 (or 3) power supplies, why not 2 (or more) e-cards?
c) _big_ servers are often connected to more than one network, so 2(or more) e-cards are not so unrealistic
(Yes, you can put 2 IP-addresses on one card)
As we speaking of servers high performance with more cards is not so unimportant (for a workstation in a pool it is overkill, no doubt).
d) You also shouldn't forget, that in the c't test they used CGI(perl-scripts) for NT for the dynamic pages, ASP will have much better results for NT.
Oh, don't get me wrong. I love my Linux machine and I love Linux, but i think i'm not fanatic.
These benchmark will help us (ahem, the kernel hackers) to improve it day by day
Frank
In any case, I'm delighted that Linux did so well against NT, even if it didn't come up on top. Given the artificiality of the test, there is considerable uncertainty in how to interpret the result.
Good upbeat attitude. Unfortunately Linux was absolutely trashed in these tests. 200% slower? That's HORRIFIC to put it mildly. And you claim it did well? Are you reading a different version of this? If I were to believe all of the armchair TCP/IP stack experts on Slashdot wasn't it a sure thing that Linux would absolutely trounce NT? Funny how reality stacks up. It is _ABSOLUTELY CLASSIC_ to read how the story is changing now. Benchmarks don't mean anything anymore. What about dynamic content! Etc. Yet not too long ago when a couple of ill researched articles put Linux on top that meant EVERYTHING.
Secondly as far as the distribution of files across 4 partitions as being a cludge: Hardly. Seldom does one department represent 100% of your traffic, or even 50%. If they DID you could implement distributed file systems (which another poster referred to as a "kludge", which is the classic term used when someone doesn't have a clue. DFS gives a substantial boost of features and functionality as far as a heirarchial share structure goes.
In any case users don't look at drive letters, they look at shares. To move a share across drives is an absolute no-brainer and has no affect on the users whatsoever unless you screw up. This means that CAD \\SERVER1\CAD is on C:, \\SERVER1\ENGINEERING is on D:, etc. But I can move engineering to E: without notifying the users in a couple of minutes.
FreeBSD uses the UFS filesystem from BSD.
Linux uses ext2, a shameless UFS ripoff.
Both have the same basic structure. UFS has
a bit more cruft, since it is older. Ext2 is
newer and cleaner, but performs about the same.
Linux uses a TCP/IP stack that is generally fast
and modern, but not optimized for SMP. FreeBSD
has almost the same, but with a bit more cruft.
If anything, Linux SMP is slightly better.
Red Hat had an invested interest to show at the test site, even if the outcome was already known. Their mere appearance with their 'experts' is directed to the Linux crowd: it makes clear to everyone that Linux is Red Hat. There are probably deriving other benefits, but since I'm not tuned into the world of Linux distributors, I have no means of knowing. As for qualified personnel familiar I doubt very much that Red Hat has th money to fund proffessionals experienced in methods of performance analysis, measurements, and statistics. Linus himself would not qualify as one.
/fafa
Interesting opinion. What specifically do think is worse? Making statements like this without justifying your stance is a lot like whining.
The kernel is not waiting for i/o it has moved on to the next thing that needs to be processed.
That's why Linux lost in this match.
Even after the failure in April(MindcraftI), they
(KERNEL developer) were still unware of the the problem. Many Thanks to MS, RedHAT, they told me where the bug was.
Lost in last game, pretended not know it, finally lost in round two. Shame!
Why the ipstack-multithread problem could not be found and fixed after Mindcraft I?
This made LINUX community silly-looking and
arkward against its hyped "Fast bug fix, Best Tech Supp"
It really annoyed me.
In c't - magazin für computer technik (German)
there is a new test:
Linux+Apache vs. NT+IIS.
Read it.
now that was an educated remark. ;-) There is no 100% unbiased testing, bat ct seems to come closer (also to real word testing setups) than others.
now that was an educated remark. ;-) There is no 100% unbiased testing, but ct seems to get closer to that goal (also including real word tests) than others.
All the servers I've worked on (in business environments) have at least 2 NICs, often as many as 4. For exactly the reasons mentioned, it reduces contention (this is further helped by attaching each to a seperate subnet).
The cost of a "global Linux infrastructure" is far from free. Software (ie NT) is not what costs money, hardware is not what costs money (although it costs more than the software if you factor in servers, fallover boxes, network infrastructure, WANs etc). The cost is STAFF. The "global corporation" I work in has probably 2-300 highly skilled NT engineers, and a similar number of Unix enginners, looking after servers and desktops.
There simply isn't a market in Linux skilled support staff, so we'd have to train them. But wait, who runs the courses?? So we'd have to hire a training company to build a custom course, then send all our support staff on it, and pay for replacements while theyre out. And of course theres the cost of moving all the data, recoding all the NT applications, retesting EVERYTHING, in parallel. Would probably end up costing more than Y2K.
Even moving our unix boxes from Solaris to Linux would incur similar problems. What Linux needs to have any major success in the corporate world is a large pool of skilled experts, with the appropriate business and technical experience. This can only come with time, MS has had a head start and has done a superb job in getting MS products into schools and unis so that graduates come out armed with all the skills they (and we) need. Maybe linux will catch up...but it will take time thats for sure.
(dons corporate MIS hat)
Okay first up. We have a single NT server running IIS 4.0 for our Intranet. Its uptime was ~18 months except for planned outages to install service packs. It serves 200k hits a day happily and all the users are happy. No reason to change.
Secondly. If we wanted to switch to a *nix, it would not be Linux. The managers dont want to take a risk. The directors hiss. We would be *much* more likely to use something along the lines of Solaris 2.6 / 7 than freeware with 'hippy' license agreements.
(removes hat)
IMHO linux is pretty cool, find the filesystem layout a bit confusing but mostly it's quite good. I have nothing against free software and use freebsd a lot for my own stuff. There is no real reason to choose linux over anything else in a corporate environment, especially considering the amount of political hassle would be involved and the fact that hardware & software costs are no problems.
Sad but true, people define the value of something by how much it costs... perhaps freebsd should be renamed $4000bsd to acheive real success...
Come on you Linux-using-bunch-of-whiners! There just isn't a day for you that some OS could be better in some way now is there?
And I is just hilarious to read how fast some users turned their sleds around. For ages Linux people have went flags high going about how fast and optimized their OS really is. Now when the reality struck you go "it doesn't matter if it's not the fastest if it's the most reliable". Please!
I just cannot believe that some people could take it this seriously. Get a life! Going around about how good OS Linux is isn't the purpose of life. How can you make it such a huge matter? Living for an OS is sure a phoemenon of the '90s.
Duh!
B
Puh-lease. I've developed for about 10 different Fortune 500 companies in the past 5 years. Never have I seen a Linux box. Ever. Please people, let's not make things up.
Linux is only free when your time is worthless ;)
I mean, let NT and Linux serving web pages and files continuously for, hum, one month. Which will require fewer reboots?
This site is so damn slow it must be running on Linux.
Personly I would probably get a honda, they have a reputation of lasting much longer, and they can be fixed up to go faster if you want.... I see linux the same way, its stable much longer, and can be fixed up to go faster (since its open source) :)
http://www.kegel.com/mindcraft_redux.html
only makes us stonger.... :)
:)
Anyone like those dragon ball z cartoons....
http://users.aol.com/suboner/Zen/
The multiple-NIC problem was solved in the experimental kernel that was available at the time of the test. According to the c't test, both Linux and NT can more or less saturate one 100 Mb/s NIC with one CPU. Since Linux' NIC interface code in 2.2.x doesn't scale with SMP, using 4 NICs don't get you anything - it performs as if there was only one of them.
;-).
Having known that, I would have expected that the Linux people either tried the most recent experimental kernel (just to show how fast Linux can get when people start fixing the bugs: at least they should have backported the fix to 2.2.10), or insisted in using a 1 Gb/s NIC - or both.
There are realistic usages for more than one NIC - routers, firewalls, combined firewall/proxy servers, etc. (ultrafast static web servers are not realistic - the internet connection for this kind of load is so expensive that a 1 Gb/s NIC is more than affordable) Linux must run fast on those with SMP enabled, too. It's really that M$ found a weak spot in Linux, and we haven't been able to fix it in Internet time. Well, we have, but the cowards from Red Hat didn't want to use that fix
I think the c't test (with Perl CGIs) wasn't really fair, either. IIS is better with ASP, Apache likes Perl/PHP. Deploy what fits best - no matter if you squeeze apples or oranges, it's all juice.
The last question people asked: are there such large sites with mainly static pages? I think, yes. A site like Geocities, where lots of people put their homepages on is quite likely to have only static pages - it's far too risky to allow all those people to put their CGI script on such a server.
... sure, that's why they do their columns on the Direct-X APIs and other Linux-related stuff. Another 'I've read one issue, now I can judge their style' reader, right?
Post it again!
WOW! LOOK AT THAT!
Linux failed a third (and forth, and fifth) time! Oh my God! We need to ditch this loser OS worldwide before it destroys us all! And embrace our deliverer and saviour Bill Fucking Gates.
MEEPT!!!!
Sure I can run my Toyota's engine at 9000 RPM and beat out Corvetts in short term races. Of course my car will fly apart and die under such operating conditions before long.
I have old Linux 1.2 boxes running as ftp/http servers and others exporting nfs filesystems that have be up for over a year. They just keep chugging along without crashing. NT has never even come close to this level of stability and reliability.
Linux is like my Toyota, it may not be the fastest, but it runs and runs and runs and has outlasted many a faster car. Oh yeah, it also does its job on 386s and 486s. NT won't even install on this hardware even though this hardware is more than fast enough for this type of network usage.
Rob needs to generate hits for his advertising buddies, so he reposts these inflammatory articles every couple of days. It's flame bait, pure and simple. Ignore it.
The German computer mag C't decided to do their own, more detailed benchmarks. Their conclusion was that the PC Labs/Mindcraft results sound plausible, but under more realistic scenarios, Linux & NT are neck and neck, or Linux beats NT soundly. Find the English version of the article here
----
Open mind, insert foot.
BSD might have performed better for this test; however now that the problem is known I am sure there is going to be a patch released shortly that will fix the problematic locking. I think these tests were very useful; they pointed out several places where Linux could be improved / tuned.
--------- Webmaster, http://www.cpureview.com and
Posted by Gates of Borg:
In you face
Posted by hurstdawg:
./'ers yelling about how this is more Micro$oft FUD we should step back and realize how good this will be for linux. Ever heard the saying, "What doesn't kill you makes you stronger?" I think that applies great here. Micro$oft is now convinced that they are #1, and Linux will never be able to compete. But, if we take this as a learning experience (I know, I know, you thought you were done with that in college... :) We can improve on our weakness', turn around with the upgrades (much faster than Redmond could) and blow them out of the water down the road, w/out them even expecting it. So I think its important we don't just think of this as more FUD, but as a chance to find out where we need to improve, and to implement those improvements.
I think before we get a bunch of
--just MHO
Posted by grabes:
I dont get it anymore. Now I will say that speed is a major part in choosing an OS. I work for a company we run all BSDI systems, and unfortunately 1 NT box. That NT box only hosts approx 75 web sites. Everytime someone publishes a page to that thing I can hear the drive grind from across the room. It has 256 megs of ram. It has what I like to call the spiraling windows syndrome. Where its really fast when you first install it and then it justs keeps getting slower and slower. However I am getting away from my main point. I have never judged NT to be a good operating system. For a server you need good Remote administation (Linux excellent NT virtually none). You need support (Linux excellent NT 1900 billme). I run Linux personally and I have no complaints I like it much better then the $1000 BSDI I also use. However older unix lovers are hard to sway from there personal choice. Now Im not a C expert, but if the major kernel developers would pull together I think we could have a kick ass os by the end of the year but thats in a perfect world. Instead be will sit here and cry about it for a few more weeks.
I've seen soooo many posting about subject where the author has made it glaringly clear he or she doesn't even know what they are talking about - and they then lye and make up credentials to back it up.
Well, I'm not lying - and I'm not about to claim I know everything... But, here's a little info about me before I start smashing myths and urban legends:
I am 22, and work as 3rd Shift Systems Operator at a Major Life Insurance firm. I'm a high school graduate, with the following certifications:
* MCP - Implementing and Supporting Windows 95
* CNA - Certified Novell Administrator - Netware v3.12
* A+ - Industry Standard in Computer Technician training. Proved proficient in MS/DOS, and basic PC Hardware
...And I have the following working experience in the following systems:
-Macintosh (PowerPC)
-MacOS 8.6
-PCs
-FreeBSD 2.2.2Release
-RedHat Linux 4.2
-RedHat Linux 5.0
-RedHat Linux 6.0
-MS/DOS
-OS/2 2.1
-OS/2 3.0
-OS/2 4.0
-Windows 3.1
-Windows For Workgroups 3.11
-Windows NT Server3.51
-Windows 95
-Windows 98
-Windows NT Workstation 4.0
-Windows NT Server 4.0
-RS/6000 (PowerPC)
-AIX 4.2
-AS/400
-OS/400
-S/390
-MVS
Whew... With that out of the way...
- RAIDS:
I've seen a few people question the reason to even test RAID performance on Linux, as it is a current weak spot. Big business uses RAID technology to support heavy load, and to break up the work of the filesystem among a group of equally capable drives. So it IS a feasible test - and just proves that Linux needs improvement in this department.- Multiple Network Cards:
A few people have said that there is no REAL life application for Multiple NICS. Linux's support for Multiple NIC cards is still premature - but it's promising (from what I can tell). I've built machines, and also maintained machines with multiple network cards - it IS a real-life factor. Another area of improvement...- Single Processors vs SMP:
Businesses will only use Single Processing machines - unless they are doing database related applications, or other heavy CPU-utilizing activity, then they will usually spare no expense to get the proper machine. Linux has primative support for SMP - but NT has better support. Linux needs to improve, so what.- Stability and Scalability:
These are both important issues in the business world. Linux currently has pretty good ground as it comes to stability. However, both NT and Linux get pounded by Solaris (so I've heard) - again, a learning experience...I could go on and on and on... But why bother...
I use RedHat Linux 6.0, as well as Windows NT, Windows 98, and even MacOS. Each platform has it's own strength and weaknesses - and any honest-to-goodness computer professional stakes his/her career on this Maxum. Don't get caught in the hype war - cause it will only lead down the road to eventual failure...
Ray D.
PS: I will probably get flammed for this - but as I have learned - that is really the only thing left of the "Professional PC Person" world... Maybe I should have become a doctor?
Posted by rdobbs:
:)
Bravo! Level-headedness in a forum of Chaos!
Posted by airborn603:
Most wouldn't expect much from an open source
OS compared to an over $500 "workstation", even
if it is made by Microsoft. Therefore, Linux
is incredible compared to NT, and all those tests
by Microsoft-owned-or-hired companies about "Linux
being slow", etc., should not affect the
GNU\Linux\UFie\Anti-MS community in any way.
* - I suspect that some people at ZD don't know
this and continue to write columns about the bugs
in IE5.
Posted by The Technical Revolutionary:
Linux hasn't lost anything. Last time I checked I wasn't running Linux because it had the fastest web server, and I could care less how fast Samba is. We need to stop chasing Microsofts tail here! Microsoft is just trying to pull linux into their overbloated Marketing/benchmark/it don't matter if it works as long as it looks good on paper world! Linux is a good, free, stable and open minded/open source operating system. Its OK if they clean up some problems in the code, but not without forgetting what Linux is all about!!
Posted by saberk:
I agree... This whole benchmarking shit is very annoying... The only thing Linux can claim now is stability. And can someone tell me who let the world know that Linux is faster than NT before testing and benchmarking it. Our (Linux) ass is kicked just because we claimed to be faster and now M$ is "proving" that it's not the case... Maybe it's time to focus more on tuning.
It was probably reposted so those who read /. during the week can see it.
-- Too lazy to get a lower UID.
I know I entered the correct URL but Slashdot keeps adding an extra space in there. Just remove the extra character and it should work.
As for the quantitative analysis, I commend the author. He didn't bash NT at all and showed the numbers. Excellent advocacy, and something more people should consider.
Lastly, for what it's worth, NT could be 4x or even 6x as fast as NT. I'm still using Linux. If speed becomes THAT critical - I'll use Sun/Solaris. NT can bite me. Oh wait. It has. 'Nuf said.
NT can be 4x or even 6x faster than Linux. I'd still use Linux. If speed was THAT much of an issue, then I'd use Sun/Solaris. Sorry, but I have never, nor will I ever - use NT for anything, especially mission critical.
The only time I'd use NT (and do) is to run Windows-only apps that I need from time to time. Quite honestly, the only reason I boot NT anymore is for 3D Studio Max, Painter and CD-R software - and even that - I can't tell you when the last time that was. I'd much rather use my SGI/Sun at that point with their native tools.
I'm not trying to bash NT here - just putting things in perspective. Great - Linux may not be THAT fast. It's still more RELIABLE. Better to have a slow site than a dead one, thank you.
Ok, the results are in. The benchmarks, when administered properly, show that NT outperforms Linux in every category.
We now need to prove the open source model instead of confusing the issue by nursing our bruised egos. I'm already seeing the "denial" posts that hark back to OS/2 days. "Well, they may have gotten better numbers in this, but that doesn't really matter." Yadda, yadda, yadda. What is wrong with you people? Can't you accept the results? Quit playing games and start making progress.
Some would love to argue with me until they are blue in the face about how this doesn't mean NT's better, etc. Fine. That is NOT what I'm talking about. The benchmark has exposed an architectural flaw/oversight. We need to fix this in order to reach the performance numbers we need to be a server operating system. So let's do it. Let's fix it. Start downloading and start coding.
I have linux installed on every one of my file servers. I will be able to fend off the criticism for now. I have faith in out community and in our talent. However, if we continue pointless arguements and pulpit pounding about how linux really is better, instead of making sure that it IS better, we're sunk!
And if you can't code, you can always test. Get involved. This is your operating system.
"Doubt your doubts and believe your beliefs."
"Doubt your doubts and believe your beliefs." -- Switchfoot, Ode to Chin
Wasn't this already submitted as a story on Friday?
This guy also needs to do some more research on some things..
A) Why would anyone use Samba on a T1 line? Samba is a file server, the majority of which are used to serve files via a LAN, NOT a WAN. LAN speeds go from 2 (10Base2) to 1,000 (Gigabit ether).
B) The speed limit on I-93 is 65. I have 2 cars, a Ford Taurus, and a Porche. Both can do the speed limit, little over every onmce and a while, EASILY.. Which do I choose WHEN COST is NOT the issue? (I am NOT saying that NT is a Porche, but the tests make the comparisons look simular)
-- I'm the root of all that's evil, but you can call me cookie..
The test was not as screwed as you made it out to be.. They had SIMULAR results. The first test didn't tune Linux as well.. BIG DEAL.. Linux got it's but kicked from a technical standpoint, hands down.. It's a sad day, but I was afraid that this is EXACTLY what was going to happen..
-- I'm the root of all that's evil, but you can call me cookie..
A) Scripting:
;-P
;-P
Please do a little more research.. ALL of these are available on NT already. Oh, and (*PPSSSTT*) Activestate is working with Perl and Perl alone. If you want bash, sed, awk, grep, and all of the toys, you should be kissing cygnus.. Oh, and did I mention gcc is there as well?
B) X thru a 28.8 Line? You'[re lying really bad if you say that it's any better then ANY sort of dial up remote admin for NT.. MANY systems can be admined via telnet. And most of NT can be admined from a web browser (As can Linux)
C) WWW Server Migration: Mute point, nothing to do with OSes here.. Just servers.. I can go from Apache/Linux to Apache/NT with no problem. Does NT now rock?
D) A sysadmin who say's 'Like Unix or AS/400's' scares the hell outa me.. Lemme set you strait.. Unix = OS type. AS/400 = Hardware Platform. As of last month, I worked at UPS, who had the largest installed base of RS/600's and AS/400's in the friggen world, so please don't try to impress me. Also, 3 years and working as a sysadmin designing a multimillion dollar ecommerce project. Doesn't add up.
-- I'm the root of all that's evil, but you can call me cookie..
The problems that people are finding have nothing to do with what your saying.. Linux hits a 200 mbit limit, along with approx 2,000 transaction limit. The has to do with task scheduling, and the NON multitasking IP stack..
-- I'm the root of all that's evil, but you can call me cookie..
I disagree.. If NT is percieved as being able to 'handle more' then Linux, then people will want the system that they feel 'can do more'. Same reason why there's so many jeeps, trucks, and sports cars. People don't NEED them, but they percieve that they need to buy what can 'do more'.
-- I'm the root of all that's evil, but you can call me cookie..
A souped up '68 Camaro will get you from NY to LA much faster than a '99 Kia will...
...but your chance of death is much greater in the Camaro.
Stating on Slashdot that I like cheese since 1997.
I'm no Linux guru and will be the first to admit that I am just a few notches over the knowledge level of a windows 9* luser... But is Samba not mostly for use in connecting Linux boxes to Windows shares? If your serving up a bunch of pages wouldn't it make much more sense to use NFS or something that is more native to the operating system for sharing between machines on a LAN? Again... I'm no Linux Guru.. just a mildly advanced user. (very mildy advanced).
- Jimbob
One thing the Linux community forgets is just how STOOOPID the average person is. And, no I won't tone down the arrogance. Here's why:
These tests suggest the equaivalent of "my car will beat your car in a drag race". Pitty, since many of us thought it wouldn't, but facts is facts.
The problem is that we have an analogous situation where those that need to find a reliable car to get to and from work daily decide which car to by on the basis of which one would likely win a drag race. Hence, STOOOPID.
Certainly, the problems identified can, and should, be fixed. They don't strike me as biggies, actually. Unfortunately, M$ will leverage this for all it's worth, and the idiots out there will believe them. Let us not get caught up trying to perform damage control. Our response should be, "Yeah, yeah, whatever -- if you want M$, then buy M$. Me, I still wouldn't touch it with a ten foot pole."
Smart money already knows that the better deal is, or will very shortly will be.
GNU and Linux have always shone when it came to helping the little guy, and the big guy will continue to throw money away. Money can buy a lot of ignorance.
In Liberty, Rene
The (yep!) english article can be found here.
Sorry, there is no (I wonder why) german page available, the results can be found in c't 13/99 on pages 186 and following.
Short summary: This test shows, that Linux+Apache outperforms NT as a webserver if there is only one NIC to handle. It also shows, that the number of processors daoesn't matter that much, when using Linux as the host sytsem for apache.
NT runs much better than Apache on Linux, when more then one NIC has to be served - which might be the result of not having a multi-threaded IP-stack under Linux.
Read ;-)
Ralph
Does this test take into consideration real-world applications? Sure, you can beat on any machine for 5 days and hope it doesn't crash. What would happen to the NT machine in 10 days? 20 days? Would it BSoD itself to oblivion? From my experiences, it would, while the Linux box would still be chugging away.
In any rate, even if NT is a little bit faster, I'd still run Linux, because in the long run, Linux DOES have the edge in overall performance. Linux doesn't crash for no reason. Linux doesn't require a reboot every x days to keep itself happy.
When I reboot a machine, it's because I want to, not because I have to.
while car can be used as exhibit, computers ussualy not. so when byuing car and money are not an idssue, then you buy car which you like better. but when buying computer and money are no issue then still you have to buy computer which can do what you want and which do it the way you want.
hany
while car can be used as exhibit, computers ussualy not. so when buying car and money are not an idssue, then you buy car which you like better. but when buying computer and money are no issue then still you have to buy computer which can do what you want and which do it the way you want.
hany
and here we have the main problem with this benchmark (and maybe all other benchmarks): benchmark tels the people that "in this particular setup doing this particular thing is product X better Ytimes than product Z". and people takes it as "product X is always Ytimes better than product Z".
hany
Multithreading could help with a single CPU and multiple NICs. A single thread could get blocked waiting for I/O, in this case a second thread could use the CPU to service requests from another NIC. Whether this is really adavantageous depends on a lot of factors including whether the IP code ever does block, and what other things can be done when one IP thread is blocked (what resources are locked during the blocking).
Dastardly
The best lesson to be learned about this exercise is that we all benefit from the showdown between NT and Linux. Regardless of who "wins" the benchmark tests, we can quickly pinpoint the weaknesses of each os and fix them. Even NT users will benefit, if MS can react in a timely fashion. This is clear support for the DOJ's case against MS, and support for open source development. I'm looking forward to the next showdown- bring 'em on!
So long, and thanks for all the Phish
I think of it like this..
Do you know any sysadmins that have switched to NT from Linux(or anything else for that matter) of their own free will? I don't.
People only use NT because their either afraid of Linux (no crime there) or pointy heads make them.
Maybe my impressions are wrong, but so far, everything I've seen and done indicated that once you've gone over to *nix, you never want to go back..
Not to mention the other(more common) end of the spectrum. I'd like to see some benchmarks for P133's with 64 megs of RAM, or 486-66s for that matter..
My little server is a cyrix 166 with 32 megs, and I only upgraded it from a 486-66 because I wanted to play mp3s...
It seems that Micros~1 inspected the linux code and found a weakness. They then publicized a benchmark that exploited that weakness. Good strategy.
:-).
Will linux ever win benchmarks against NT if Micros~1 just benchmarks NT against linux's weaknesses? They will just find another one.
The good side is that Micros~1 is actually funding linux development by finding weaknesses. Perhaps linux will one day be perfect and there won't be any weaknesses left
The bad part is that linux may never win any benchmarks, since only Micros~1 can afford to test them.
Eg., why haven't we seen a gigabit ethernet benchmark?
support gun control: take guns from cops
http://cs.alfred.edu/~lansdoct/mstest.ht ml
Kinda puts the whole shebang into perspective. Watch out for the twist of irony at the end.
Regards,
Regards,
-scott
Again with the free beer stuff? If NT is faster -whether because of performance or cost of training is lower because it's easier to use- then it saves money. Time is money too.
---
Seriously... This was supposed to be a fair test, and until the RH team that was attending (and configuring the Linux machines for) the test says otherwise, we'll have to accept that.
However, I think a Mindcraft III would fare _much_ better now. The fixes from Mindcraft I didn't come in in time for Mindcraft II (see an older article on Slashdot), and now the RH guys found even more things to fix. In addition, as the article pointed out, Apache is getting better, too. (What about Boa, or any other single-threaded web server, BTW?)
Sorry, folks, this was a fair test, at least when it comes to biasing. However, as I've said, Linux was a bit unlucky with the time and date. (Microsoft, on the other hand, had already made their improvements, based on the data from the previous tests... Guess they've invested a lot of money in fixing it quickly...)
All in all, however, this was a (minor) win for Linux -- the results was nowhere near Mindcraft I, even with Microsoft's new improvements.
/* Steinar */
(This comment is of course GPLed.)
To me, multithreading here just looks like a `buzzword', to explain something more difficult to the end readers.
:-)
Please get this right: A well-written, single-threaded function will beat (or equal) an equally well-written, multi-threaded function in _all_ cases, simply because somebody has to do the switch between threads (processes) in all cases. In the multithreaded version, the OS/CPU has to do it. _Multithreading_ _doesn't_ _mean_ _extra_ _speed!_ (That's why I wrote my single-threaded FTP daemon, BetaFTPD (search Freshmeat for it if you're interested). Now, just had to pull the plug for my own (GPL'ed) `product'... Don't make that affect your judgement, please...)
Now, got it right?
/* Steinar */
(This comment is of course GPLed.)
Yes, of course, but as far as I understood the article, this was true for the uni-CPU version, too.
BTW, isn't 4 Ethernet cards a bit unrealistic? Wouldn't a single gigabit card be much better?
/* Steinar */
(This comment is of course GPLed.)
Remember that every process (as you referred to, threads are only lightweight processes, also on Linux) takes up additional memory. Less memory for cache, or ultimately more usage of swapfile.
/* Steinar */
(This comment is of course GPLed.)
>> you cannot remote administer NT (Im not talking about fast connections here, (where you could use VNC), try to administer NT over a modem line. Good luck) > I've used both PC-Duo and Carbon Copy to do remote NT administration just fine over a 28.8 line. Hmm.. like you'd get those thru a halfway decent firewall. I have to go down 5 floors to admin our servers from the console because no remote admin tools were included and being a lowly System Administrator I wouldn't be able to get funding for a remote admin tool let alone get it through our firewall.
Sorry but NT lags sooo... far behind on this one, I can use X thru secure shell from a trusted host with x.509 certifate authentication to my linux box - in fact I haven't touched my linux boxes keyboard since I moved it into a different office.
>> once youve made your decision to use IIS, youre completly stuck when it comes to changing to another type of webserver, or sometimes even when you want to transfer sites from on IIS to another I've never had any problem with that... the MS tool works well, and its not like on a Linux box you can move seamlessly from Apache to Zeus, for instance. I have - it can (not necessarily always is) be a pig, any move can be damned horrid, but in my experience so far UNIX (linux/aix) stuff is simpler to migrate because I am in more control.
To put this in context I am working on a multimillion dollar e-commerce project using NT,AS/400s and RS/6000's. If we had used unix or as/400's instead of NT/intel it would have been completed by now and more scalable to boot.
I have been administering unix for several years and have 2 years + 1 year industrial placement of my BSc Honours Degree in Computer Systems & Networks under my belt so I know what I am talking about.
A.
Intranet/Internet Developer & Linux Advocate
Why was RedHat chosen? A proven stable Linux server setup by a Linux professional would have been a fairer representation. In my experience the workhorse of Linux has always been the Slackware server. Even Caldera or SuSE would have been a better choice. The NT machine was tweaked by NT professionals so why not at our end. RedHat is not exactly known for it's installation ease and stability and usually requires a higher degree of expertise so the test was handicapped right from the start.
-- Ted tsikora@powerusersbbs.com
How can you say that. The library differences alone make or break it. Let's see we have libc5 5.3.12 vs 5.4.46 / glibc 2.07 vs 2.1.1 Bind v4 or bind v8 Where all the apps compiled for the current libraries, dynamic or static? Static libraries consume less memory hence are more effecient but are larger. (FreeBSD one of the proven performers is a-out.(2.2.8)) The amount of differences are staggering and their effect on performance absolutely critical but that's the beauty of Linux. Build it your way.
-- Ted tsikora@powerusersbbs.com
I would be inclined to agree with you, if it weren't for the fact that the PHBs and bean counters are trying to pick between NT and Linux.
If the management types are asking the question "NT or Linux", then why shouldn't the Linux community be interested in any data that could sway somebody's answer?
The "right way", of course, is to pick the OS and software that does exactly what you need, etc. etc. Unfortunately, for arbitrarily complex real world situations, with arbitrarily complex management, personnel and customer pressures, there isn't always the time to find the "perfect" system. So, instead people rely on benchmarks, press reviews, word-of-mouth, etc. And also, unfortunately, management often trusts the word of the outside world more than the word of its own employees. (Sometimes, there are seemingly rational reasons for doing so, such as "suppliers X, Y, and Z provide more solutions for platform A than platform B, which is more important to me than the fact that Frank prefers platform B because 'he likes it better,'" neverminding the reasons why Frank might like it better.) And so, monitoring the press and actually responding constructively (eg. fixing performance problems, etc) is a good thing.
The key, of course, is to keep sight of what matters. Lets not get trapped into optimizing things that don't actually matter. For instance, optimizing Apache for utter speed while compromising stability would be a bad thing. Or worse, bloating the kernel with benchmark specific tweaks, etc. (After all, isn't that how Microsoft attains its legendary (lack of) stability?) Remember, while these sorts of highly-publicized benchmarks may point out some weak points, they also could just be serving as chaff to block the real threats from our radar screen.
Perhaps that was the idea you were trying to get across, AC?
--Joe--
Program Intellivision!
Isn't this the exact behavior that SCSI disconnects are supposed to help?
I'm pretty sure the kernel doesn't block waiting for the I/O to complete. It issues the request to the device and puts the blocked task to sleep while it goes about handling Everything Else TM.
Now, in the case of IDE, if both drives are on the same IDE chain, you're going to block waiting for the device driver. If they're on different chains, then they can run independently.
--Joe--
Program Intellivision!
When people choose to dump their MS OS and go with Linux, I've never heard "because Linux serves more pages per second than NT" given as a reason.
The reasons are usually one of the following:
Better Stability
Tired of being locked into MS-only solutions
No license fees
Doesn't need an expensive hardware upgrade
No annoying features like animated paperclips
Linux doesn't pretend to be smarter than you, and then fail at it.
No registry which causes more pain and grief than it actually solves.
It doesn't really matter than NT beats Linux in this test. Sure we should fix the problems, but NT is still NT with all the liabilities which it has become famous for.
Of all the comments I've ever posted, this is definately one of them
Both the PC Week and Mindcraft test prove NT can outperform Linux at static webpage delivery. But this is not interesting to me. Most website development I'm involved with involves dynamic webpage delivery.
Without a doubt, Linux (or any Unix-like system) would squish NT into the dirt if they tested dynamic content generation... especially if they did it with CGI. By design, NT is process-heavy and has considerable more overhead to deal with when forking processes (i.e. CGI scripts).
While I agree the Linux and Apache coders have some work to do to get static delivery up to and past the performance of NT, I'd like to see some numbers for dynamic content delivery (the reason I use Linux and not much else).
But the main point is the overall impression. It's a much more differentiated view than the other tests. Particulatly interesting are IMHO IIS's problems with delayed requests (3 sec). This seems to be the downside of the threading approach.
Could you please be more precise?
C't is always biased in favor of the more advanced technology (guess what the "t" in c't stands for), and that means they are open to BeOS, Linux/*BSD, (in earlier times) OS/2, and, as they try to be comprehensive, MacOS (X and below).
As for the c't tests, I've yet to find someone who proves serious factual errors therein.
On the contrary, c't tests tend to be very sophisticated and neutral. No wonder, as c't employs the best computer tech journalists in the German-speaking world, probably even world wide.
Maybe you'd recall
PIII proc ID software switch by processor guru Andreas Stiller,
the SoftRam scandal, where everybody from APC to ZiffDavis hailed this as an ingenious way to get cheap 'Software' RAM, until c't examined that the company had just put non-functional (unused) code to the standart MS libraries...
Well there are many other things I could mention, but in general it's save to say c't is one of the best computer mags worldwide, certainly the best I know (and I know quite a number of US/UK/french ones as well).
It's well known that ext2fs is slow, unefficient,
non-scalable, and not very fault tolerant.
I'm not surprised at all.
What is surprising is that linux ip stack which was supposed to be very efficient, was proven to be very uneffective.
It's interesting how would FreeBSD deal with such load, FFS or vinum are much faster filesystems, and BSD stack is much more effecient. We'll see..
--Roman , remove spamless. to mail me.
I know. i have 2 servers right here with uptimes
in the hundreds of days. I only recently noticed that they were only using 64 MB when there was 128 MB of RAM.
In 2.2 Kernels i can testify that problem is no more.
Ok, now you know how the BSD folks feel when Linux gets the limelight even in circumstances where a BSD is the technically better solution.
There's no real reason that Linux should be in the front of everybody's minds, and BSD hardly heard of at all. It's just marketing; Linux folks are a lot louder than BSD folks.
cjs
The world's most portable OS: http://www.netbsd.org.
I like how the exchange went between Steve Jobs and Bill Gates just as Jobs realizes that Bill had stolen from the ideas that Jobs had stolen from Xerox:
:)
Jobs says: "Our stuff is better..."
Gates replies: "Your missing the point. THAT does'nt matter..." (something like that anyway
Cool irony huh?
I have heard that Windows NT uses a BSD stack. That technology has been around for a while. Now Linux networking code derives from a similar background. How can their be such a performance different if Bill (gates) snagged BSD code? Could someone please explain this to me. I may not have my facts straight. Also, NT uses 7 mb MINIMUM just to load (according to NTLDR) + cpu time to run the gui + disk access time for virtual memory. Is their gui so well written that it minimizes the usage of such system resources? Well, I am a computer tech & I have seen what NT can do. I optimized an NT system where I work & later, I installed 95 on the same system & optimized it (disk cache sizing, vm adjustments, path/file cache, etc.). You would think the Pentium 200 I set it up on was a Pentium 100 running NT (32mb ram/3.2gb disk). I know NT should probably have more disk & double or quadruple the ram, but why? Why do you need such a beefy system just to run the stupid operating system. Why a gui? It will just sit their most of the time doing nothing but wasting resources. How can NT still come out on top? I want some answers.
This was probably a pretty unbiased test.
Now, instead of whinig and bitching, lets get serious and fix the problems. This is not the last time "Bills Boys" beat us, but if we keep on whining instead of fixing the problems we may just give up here and now! We have to remember a few important things;
1) There is no such thing as bugfree S/W. Linux is good but not perfect.
2) Linux and OSF/FSF has com a long way in a short time. Let's not loose focus now.
3) "Bills Boys" will of course look at this and try to be even better where Linux is deficient.
4) Linux is _way_ more stable than NiceTry, let's open that gap up even more by removing even more bugs.
5) Whining and crying foul will only hurt us all
Microsoft would never have pushed for these tests if they weren't 100% sure they would win.
...) for months? Or performance tests in an area where linux wins today?
They are possibly doing their own tests with linux in order to find scenarios where they can win. Then they announce a test, possibly inviting linux people to tweak the linux machine in order to gain credibility. Then they can say "They worked hard at it - but we still won!"
This isn't all bad, as we get hints on what we should improve. But we should set up tests ourself too, balancing the opinion by showing off some real-life scenarios where linux wins already.
How about a stability test with a machine that runs many different services (file/print + database + dynamic web +
People who want a purely static webserver now knows to use solaris or NT. Some other benchmarks might let them know when to use linux too. Red hat & other linux vendors could surely fund some tests themselves.
loser. After having put up with FreeBSD for 2 years and observing how second-rate it was in functionality, performance, and stability to linux, I was glad to replace it. The few recent comparison benchmarks I have seen confirm that, and regarding your signature, no, linux has many ports than netbsd!
tcboo
many _more_ ports than netbsd...
Even your web server is offline!
tcboo
Absolutely! They can't possibly compare the monolithic NT against an almost infinitely configurable system. "Oh, ip stack wasn't MT? Just replace a few library calls and recompile with the best MT library" (RH's patch already available!). Or egcs, or pgcc, or other commercial compilers already available. Recompile a minimal kernel, turn off extraneous processes, pin files and executables in memory...
tcboo
well, guy, it's not that we are blind to M$'s good products. For most of us, we have had to _endure_ these second-rate, poorly designed excuses for "productivity" software until we found the superior solutions available in the open-source world. You obviously have no idea what "scripting" in unix is about if you think VBscript even qualifies.
tcboo
Of course MS is going to win these things. I'm sure they have a team of coders who concentrate specifically on making NT perform well in benchmarks.
What people are failing to note is that NT costs hundreds of dollars a seat, where Linux is free, as in free beer. A global Linux infrastructure for a corporation costs $0. Of course professional support is much more than free, but the fact that the code is so accessible means that an organization can have it's own on-site support staff with access to every bit of information they would need to fix the problem. There's a lot of money to be saved in using Linux and people will figure it out. When they do NT's going to have a tough time.
-Rich
Now, what corporation would use a loaf machine like yours to do corporate web serving and/or file sharing? If you are reaching this far in order to refute the truth, why don't you mention that NT won't load on that PS/1 you have sitting in your corner... or on that 8088 with dual bad-ass 5 and a quarters?
OTOH, it is feasable for a startup or smaller corporation to be using a Single CPU Pent III with a measily 256 MB of RAM.
www.jackasscritics.com
Who cares what distribution they used? What matters is the versions of the kernel and servers. Since they had people from Red Hat there to tune the system, it made sense to use the distribution that they felt most comfortable with.
Distributions only matter in performance if you're not going to take the time to update and configure your software, or if you're not going to disable irrelevant services only provided by some distributions.
Arguing over distributions for benchmarks is like arguing over Gnome vs. KDE when you're not running X.
http://www.kegel.com/mindcraft_redux.html
(Found at LWN.) This page is full of technical analysis of the Linux kernel and Apache, explaining a number of performance problems that the benchmarks brought to light, as well as solutions for many of them.
It sounds like a repeat test in another month or two would bring things even closer. With khttpd, we might even win.
One of the most important points why more and more companies choose Linux/Apache
has nothing to do with speed or price of the OS or the webserver. The really
expensive part when running webservices are administrative costs. NT & IIS may win here
for *very* small sites, and UNIX wins when it comes to housing more than a few dozen
sites - enough to make NT and especially IIS unmanageable.
* NT has no useful scripting, Linux has everything you can ever need
* you cannot remote administer NT (Im not talking about fast connections here,
(where you could use VNC), try to administer NT over a modem line. Good luck)
* once youve made your decision to use IIS, youre completly stuck when it comes
to changing to another type of webserver, or sometimes even when you want to
transfer sites from on IIS to another. Microsoft has a tool to do that job, but -
guess what - it crashes on even the smallest problems. Apaches configuration
files can easily converted to another text-based config - use sed, awk or whatever
youre use experienced with.
* if something goes wrong with IIS, the event log will contain such useful error
messages "could not bind instance XXX. The data is the error code. 43 00 00 6c".
* IIS is a hell when it comes to logging. All logging is done asynchronously, so your
only chance to see whats going on is to wait a few minutes for IIS to sync() the
logs. Really a pain when you want to study the logs...
* Sometimes under NT, the MMC console simply is stuck. Then your only chance to
get it running again is to restart the system, simply logging in as a different
user does not help. Very annoying.
* Any finally, those beloved situations where those windows popup:
"Your system is running low on virtual memory...". When you check the taskmanager,
it will show you that neither applications nor the system itself seems to use
that much memory. Again, your only chance is to restart the system.
Thies tests remind me of when PCs and Macs were being tested. The tester determined who would win. If it was a Mac mag the Mac would win if it was a PC mag the PC won. It's protecting ones investment in a platform. Microsoft (The evil empire) dosn't have to approve in the slightest. I suspect that even if Bill Gates himself sent them nasty letters they'd keep doing this sort of FUD.
The way it's done is easy. You criple it in some major way that the public isn't going to notice.
In the Mac vs PC tests it was done by pulling/disabling processor cache. It can also be done by disabling features in the Linux kernel (not hard..) or enabling features that might interfear with the preformence.
An easy way to beat out NT is do pritty much the same thing to NT and optomise Linux. Same results.
I say let the FUD fly. This fud is aimmed at the techno savy. Like telling a car expert a KIA is better than a Lamborgini and proving it by putting a lawnmower engen in the Lamborgini. It sounds desprate. Would you believe someone who regularly make false clames?
Ok it's a matter of catching in the act.. consider them cought.
A reminder to my fellow zellots... DO NOT FLAME ANYONE... and DO NOT send e-mail to people who issue FUD. Leave them ALONE.
Those that need to know, know better...
and if they can't tell the diffrence between fact and FUD then we are better off with them running NT becouse a bad admin will crash a system faster than anything and that makes what ever os they are running look bad.
I don't actually exist.
It shouldn't be to long before someone fixes the bottleneck. I doupt thats the real reason however.
Why? All the FUD, the history of thies tests and the general addatude of people who want Windows NT to be thought of as suppereor to Linux.
Microsoft has there own Zellots and some of them are in the PR department.
Linux is a good Os and so is Windows NT.
It's just not good for the same reasons
I don't actually exist.
> 1.They cheated.
I have come to expect that ANY such benchmark where someone comes out better than someone else someone cheated.
In a realistic test NT and Linux should come out equal.
Back during the Mac vs PC debates PC and Mac mags would do benchmarks a lot and the winner depended on who ran the test. A lot of the ones I saw the losing computers cache was disabled resulting in a major test bies.
I can't see anyone doing a benchmark with out first having a stake in the results.
I don't actually exist.
It's way to easy to fudge benchmarks.
They seem to rely on defects in Linux. It is a good idea to fix the defects but don't expect that to improve our scores.
Anyway speed isn't going to keep you on-line. Reliability is.
Anyone want to do a reliability bechmark?
I don't actually exist.
I don't buy the "expert" and "cost of retrainning" arguments.
Un*x experts have been around for a long time.
The money you save in not trainning NT admin and not testing NT servers dosn't cover the losses caused by untrainned NT admin and badly configured NT servers.
The time and effort put into Linux is the secret behind reliable Linux servers.
The free beer argument holds up much better...
Some times free beer is a bad thing...
In the end it's what works best for you...
It was true a long time ago and it's true today...
I don't actually exist.
NT is faster than Linux in this benchmark. It is far from real-world. In the real-world, Linux has proven itself to be faster than NT, repeatedly for many people. I suspect it was the 4 NIC cards that gave NT the advantage. In the real world, a) you would not be serving static web pages over 100Mbit ethernet, and b) you would be using 4 uniprocessor linux boxes, which would be much faster, much much cheaper, and more reliable than one big monolithic machine. Its funny that MS chooses to use a monster machine instead of a cluster of small ones, which is what they advocate when comparing NT to Solaris. This just shows that MS will pay for benchmarks that NT will win, no matter how contrived they are. MS wins by marketing to managers, who know nothing about servers, and believe this "real-world" benchmark. Techies everywhere got a good laugh from it.
Facts:
* One condition of the re-test was that the Linux system could not use patches available after some date back in April before the first test. Therefore it could not use the patches that fix the TCP/IP bottleneck that are already in the 2.3 kernels. Microsoft, however, were allowed to use patches they made to NT after, such as the NTFS fixes.
* They still use Apache even though everyone knows Zeus is faster. They make a claim that it was "not much different" but there is no data given to back this up. Even Apache Group ppl will tell you that Zeus is much faster, since Apache is designed for flexibility and not raw speed.
* On that subject, the IIS server used has a fast path to spit out static pages. Apache does not.
So pretty much we have the same situation as the first test - highly tuned NT beats poorly tuned Linux. Except this time, as many people feared, the results have some kind of "stamp of approval" on it since Redhat were involved. Even if it was some guy nobody has ever heard of before (which, judging from past experience with Redhat, generally means some college kid on summer vacation). I can see the headlines now: "College Kid can't tune Linux as well as professional NT engineers from Microsoft can tune NT"
Oh, hang on, that would require truth in journalism.
Matt
I think you're right. Microsoft can rapidly improve.. when they're scared. With Linux nipping at NT's heels, watch for Microsoft to attach Linux head on, without actually admitting it. Throughout all these Mindcraft and PC Week benchmarks, Microsoft has been particularly quiet. I don't think they even want to admit Linux exists. They'll let their henchmen do that.
Windows 2000 is so bloated and overdue, I doubt Microsoft can change it much to address Linux's "thin server" attack. So watch out for Windows 2001. Displeased with Windows 2000 looong schedule, Microsoft has said they want the next NT dev cycle to be much shorter. (This sounds a lot like Linus' wanting Linux 2.4's dev cycle to be a lot shorter than the 2.5 years spent on Linux 2.2) I bet Windows 2001 will ship 1.5 years after Windows 2000. I'm sure it will be a (somewhat) thinner beast, ready to pounce. The next few years of Linux should be very exciting.
cpeterso
Admittedly, I have not used FreeBSD or Linux SMP, but I have read about both operating systems. Doesn't FreeBSD use a single big lock for the entire kernel ala Linux 2.0? Linux 2.2 has kernel threads and finer lock granularity. Does FreeBSD have kernel threads, too? I heard someone paraphrase this as: "Solaris kernel threading is 2-3 years ahead of Linux and Linux is 2-3 years ahead of FreeBSD".
Does anyone here have first hand experience using SMP on FreeBSD and Linux? Any problems or preferences?
cpeterso
By bloated, I didn't mean slow. I just meant that for a "lean and mean" server OS, it sure includes a lot of extra packages, like IE5 or Windows Media Player or Pinball or 3D mouse pointer shadows. I don't think this stuff is going away, but I think Microsoft needs to address this issue if NT is going to be "enterprise ready".
cpeterso
Do you think that many people reading /. would really hammer on the Linux server? They would try to crash the NT server. Of course, if they ignore the Linux server, it's throughput measurements would be lower.
cpeterso
People won't miss playing videos on their servers. Also, IE5 does not replace Windows Explorer. Otherwise, Windows 2000 would not ship both EXPLORER.EXE and IEXPLORE.EXE.
cpeterso
Look, alot of you are complaing shouting fud and all. I non-biased when it comes to ms and linux. But lets look into what the claim to be the problem and not try point the blaim some where else.
They stated that the ip stack isn't the greatest. We all know the BSD's have great stacks, why don't we look into this.
Also linux isn't the greatest at smp also. BSD is good at this also.
Instead of trying to go own your own totally, you must look at what examples that we look at and try to base our fixes on those examples.
Instead of whining, do some research on some of the kernal that do things better and try to figure out how to incorperate them into linux, so the next time we want to race, we aren't using a porsche with diesel fuel trying to race a mini van with with 120 octane racing fuel. The mini van will win, becuase it's using the correct gas, but isn't stable to run for long. The porsche wont even start up, if it did, it wouldn't go anywhere.
So please stop all this bitching, becuase you are just waisting valuable time which can be spent in the redesigning and fixing our problem that we have in linux at the moment. Linux relies on the users to make it a success, not just a select few.
I ate my tag line.
I ate my tag line.
-=Ellis (D)25=-
This article is fabulous! These guys get an A+ in my book for objectivity and for being informative. They basically tell the reader every little thing they did and what the effects were, regardless of who's ox was gored. What's wrong with the American rags that they can't be as technically complete as C't? Too many hidden agendas? Too worried about whether they're violating someone's NDA?
... and Linux came out marginally on top. Conditions were a good bit more real-world, and the article is pretty well-written. Read it in English at http://www.heise.de/ct/english//99/13/ 186-1/
(Info found at http://www.lwn.net/daily/)
Have you got some proof in the form of unbiased figures to back up these claims about *BSD?
If so, please post them (or a link to them), since I'm very curious about this...
If there is one thing that Microsoft has done well in the past, is that it has adapted quickly to threats. It looks like they are again changing their ways, this time to combat their demize at the hands of OSS and FS.
The big difference this time, is the fact that they realize that they can't win with marketing alone... They can only win by improving their own product.
In my opinion, this is a welcome change. It show that competition in an open market will eventually provide better products.
Who knows, perhaps we will one day see an OS from Microsoft that is free and open (not likely, but not entirely out of the question)
For now however, I will keep using Linux for my own use because it fills my requirements (not because I think its the best of the lot, that is inconsequential)
There's an interesting article I read off LinuxToday which analyses the results of this benchmark.
NT smoked us.. but unless bandwidth gets a whole lot cheaper REAL fast, no one will care!
check it out:
http://cs.alfred.edu/~lansdoct/mstest.html
--- Tao
see subject
I am sorry, but I humbly disagree. IMHO, they got it right this time, by testing the most optimized HTTP server on each platform: APACHE for Linux and IIS for NT. That works great for me, since one would most likely not use APACHE on NT due to the lack of optomization vs the linux version. Just my 1c.
Real-time Collaboration Consultant
++ Say to Elrond "Hello.".
Elrond says "No.". Elrond gives you some lunch.
Give a faremer, and a city boy a fast car each. Tell them to go home and se who can drive fastest.
The farmer has a wery bad road, and cant use the car home.
The city boy has the highway to drive on.
You chold give two different companies the same amount of money
Then they shold set up the fastest server. One company
with NT the other with Linux. One of the companies
have to pay some $$$$$$ for the licence, and the other
company can by a huge cluster. Then test mail, web, dns and all the stuff.
Then the system cold be tested for 5000 users. That will be a huge lisence fee to pai for.
Go here
The power of the Linux community is the ability to mobilize and focus on problems in a way a traditional company can't.
These results will simply focus the Linux development effort to remove these issues.
If anything, Microsoft should be running scared of the next compairison. The Linux coders were just shown the plan on how to beat NT in the next round. All the Redmond folk can do is hope to stay ahead.
nuclear iraq bioweapon encryption cocaine korea terrorist
While what you say might be true for a single CPU, it certanly isn't the case when you have more then one, if you only have one thread, that means that only one of your CPUs can work on your program at a time, and in this case, only One of the 4 CPUs was running Linux's IP code, with the rest going to waist...
_
"Subtle mind control? Why do all these HTML buttons say 'Submit' ?"
ReadThe ReflectionEngine, a cyberpunk style n
Now if only the Linux Community can learn to implement this interpretation of benchmarks ....
That is to say: folks, benchmarks are our friends, EVEN IF they're biased !! So what if the NT box uses an Apache server ... I'd like to see WINE use IIS, just to say that we've tried it -- wouldn't THAT be a hoot !!
"He who questions training trains himself at asking questions." - The Sphinx, Mystery Men (1999)
Don't you see what is going on here..
... FREE! (Free Beer!) :)
We didn't loose anything.. We have now a (How many billion dollar company?) trying to protect their reputation. Microsoft is on their heels and if they think this will stop the community and tarnish our reputation.. wrong. The fact of the matter is Linux is a much more managable system than Windows from the code standpoint. How much resources are they putting into these benchmarks to keep beating us.
Obviously we are the under-dog, it might take 2 years before we can beat them in a benchmark... but who cares. The world *loves* an under-dog, and we are here to stay.
They are not going to crush us with a benchmark, they are going to crush themselves with the great publicity we are getting.
Any fool can look at the benchmarks and see that in reality, even if Linux isn't serving as fast as NT (According to the benchmark results...) Hey, it's fast enough. Everyone knows that the system is VERY stable, and also
PAHLEASE!
..
/., hey.. more power to them.. more power to rob.
:)
Geez dude, get off your soap box. My lord..
So what if Rob takes money from Microsoft, I would too if I could. I appreciate the time that Rob has put into this site
If Microsoft wants to sponser
(I am having a flashback of the reactions when winamp was bought out by AOL...) More power to the sap. They are doing a hell of a lot better than I am.
In reality, I can make a choice.. I can choose Windows or Linux.
Doesn't matter who is sponsering who.. who is using what. I don't care if linux never becomes more popular.. I just love the thrill of running on a new operating system. The day it becomes ultra popular/defacto OS.. I am sure that something new and innovative will come out to replace it. That is the great Circle of Life!
in a proper scientific experiment, there is only one variable, in this case it was (read should have been) the operating system. However, they also added the additional variable of IIS. They should have used Apache on both, then it would have been a proper comparison...
Opinionated Law Student Strikes Again!
for those of us with mice
http://cs.alfred.edu/~lansdoct/mstest.ht ml
Gotta love numbers. I think this puts more a real world spin on the whole thing.
To paraphrase metaphorically....
My New Tech car goes 5000 MPH.
Oh yea, my Free Open car goes 2000 MPH.
But the speed limit is 55.
Yea, but who obeys the speed limit.
O.k, but the ROAD breaks down at about 500 MPH, so where does your car go after that?
And BTW, how often does your car break down? How much did it cost?
..and the final sentence:
I thought everyone flew planes in the real world?
+&x
Oh, you wanted one that was actually connected to something? :)
Save the whales. Feed the hungry. Free the mallocs.
I propose a new benchmark based on the /. effect. I'm sure someone else has already said this but I just thought of it and can't recall having seen anyone else post the idea.
/. to generate over 5,000 hits/second aren't there?
;)
We get M$ and RedHat and anyone else to set up servers with the same content on a known, solid T3 connection for each box. Then we post the URLs to slashdot and see which box can handle the load. According to ZD's numbers the machines should blow. There are enough people on
BTW, anyone have a T3 we can borrow?
Do really dense people warp space more than others?
Here's what the second benchmark showed:
:^)
1. The Linux Advocates Were Right
Mindcraft *did* seriously mis-configure Linux in the first test.
The Linux/Apache peak performance in the second test was approximately 50% higher than in the first test.
More important, the disasterous collapse of Linux/Apache at higher loads, that occurred in MindCraft's first test, was nowhere to be seen in the second test - Linux/Apache performance remained high as the load increased. The performance drop-off in the first test was caused by MindCraft's mis-configuration.
2. The Anti-Linux Zealots Were Wrong
Linux advocates did *not* oppose the first test simply because NT beat Linux. The opposition was based on valid concerns about how the test was run - concerns that have been born out by the second test.
There is little serious opposition to the second test, which is generally considered fair (within the limits of the benchmark). In fact, the knowledge gained from the second test has been welcomed by the Linux community, who look forward to the performance gains that will result.
3. NT/IIS Beat Linux/Apache - Not That It Matters
IIS on NT *did* achieve a higher benchmark result than Apache (or Samba) on Linux. But, as many have pointed out, the conditions of the benchmark are highly artificial. In the real world, where there is a greater mix of activity on the server, Linux's virtues in the areas of stability, task management, and I/O performance would play a greater role.
As some have pointed out, when you're shopping for a reliable delivery van, the fact that it can be beat by a dragster is of little consequence. Or, to use another car analogy, by pouring on the nitrous, you can beat any other car on the track - for one lap (before burning out your engine).
4. Linux/Apache Performance Was Excellent - Not That It Matters
As others have pointed out, the Linux/Apache performance on one CPU was enough to handle the load of *twenty* T1 lines. But again, the test is too artificial for that to have much meaning.
5. That MindCraft Guy is a Whiner
There are always jerks on every side of every issue. For him to pick out the most obnoxious things said by some Linux supporters, and suggest that they mean something, is childish. Those mouth-offs don't represent the Linux community, anymore than Ballmer represents the...NT...oh...wait...
Yea but it's not difficult to find out what web server is running by telnetting to port 80.
What if the tests are right?
All it means is we have fodder for the next revision.
It's a thankless job, but I've got a lot of Karma to burn off
I'm sure that NT doesn't run as long as linux would, and you've got to count in the fact that a 5 client NT server liscense probly costs somewhere atound $400 ... while you can do anything you'd like to linux, and use it in any manner for nothing.
-neil
Well, not exactly. If I read correctly, they split up the data being served into four different file systems so as to avoid an NTFS bottleneck. If this is true, I'd view this change as more of a kludge than a tweak.
To see the kind of boost they got on the benchmark in a real world scenario, you'd have to arrange things such that users use each partition equally. My sense is that this would be nearly impossible or at best would make you very unpopular ("Fred, too many people are using 'L:'; we need to load balance the NT server. Can your team arrange to distribute their CAD files between 'L:','M:', 'N:' and 'O:' so that they use each drive equally all the time?")
In real world scenarios, rules like the 80/20 rule, or Zipfian distributions (v/n approximately a constant, where v is volume of use and n is numerical rank by volume) are the norm. Unusual peaks in usage tend to be even more skewed. In each of these cases it is unlikely that you would see any real world improvement from the changes the MS people made for the test.
Of course I only read the summary of the changes; maybe somebody more familiar with the mechanics of the actual test could set me straight if I'm wrong?
In any case, I'm delighted that Linux did so well against NT, even if it didn't come up on top. Given the artificiality of the test, there is considerable uncertainty in how to interpret the result. Given this admitted slop, it looks to me like Linux performance is in the ballpark on NT's home turf (not to mention a great way to make good use of surplus 486s you'd never run NT on). If we get a better, multithreaded IP stack out of this to boot, great, but I wouldn't sacrifice stability and security just to be able to catch up on a benchmark machine with 4 NICs.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
This time I will say what was learned from the so called second round of benchmarks: 1) It showed that Mindcraft didn't know what they were doing. 2) It shows that we need improvements, and it was very helpful in stating these facts. I love Linux, but I won't turn away from its problems, because if I or any of you do, we are sentencing Linux to be a second rate Operating System, when it really doesn't belong in that category. Its great to see people talking so great of the operating system you have accepted as the operating system that has almost everything. But I know that it hurts when you see it being put down in the wrong light, but the thing one has to realize is that some things are true and some things are lies, and you have to be able to know tell which is which, so if you know how to program, pick up the source code is a check it out, if you find a better way to do something make the changes, change it and see if it puts out the proformance you want, and send it to Linus, or who ever is incharge of that peace of code. Thats all I have to say. Oh yeah Debian Rocks!!! hahaha
I don't know what life is, but no one gets out alive...N
Did the benchmarkers remember to add append="mem256M" to their lilo.conf? I know it sounds silly, but sometimes it horrifies me how many servers are running with only 64M of memory because configuring more memory isn't so obvious.
Or is 64M no longer the default upper limit of memory in the 2.2 kernels?
It baffles me that you'd have to go the extra step to add that line to detect extra memory in the first place. Can anyone provide an explanation?
This is a really interesting spread of comments on this matter. We have on the one hand the Linux advocates (which I will happily count myself a member of). Opposite us are the NT advocates. Both groups make good (and bad) points, reflecting a range of reasons from excellent and well-spoken, to extremely close-minded. In short, it doesn't matter _why_ NT beat Linux so much as what it means to have this kind of attention. Yes, the hardware was more conducive to NT than Linux. Is this unfair to Linux? Nope. Does it mean that NT is better than Linux? Nope. It only means that NT outperforms Linux within the limited parameters of the test. Additionally, some limitations within Linux were uncovered that can be improved on. This is excellent news! All of this states quite firmly (to those who choose to look at the big picture instead of a few numbers, however captivating) that Linux and NT can be seen on a comparable scale. Not bad for an OS that scales well on a 386 or higher. Keep in mind just how flexible Linux can be, as well as how committed its development and support is. It's never shameful to be beaten in a contest. It's only shameful to refuse to improve yourself for the next contest. Peace.
--
ups, :-\
this would really be horrible!
not only, that linux doesn't provide any information (100%slower) but it also sucks the other 100% from the user connecting to the side
-- all those moments will be lost in time like tears in rain.
Maybe instead people should look at
http://www.heise.de/ct/english//99/13/186-1/
Which describes a rather different set of tests.
you would be using 4 uniprocessor linux boxes
Or you'd be using 5 boxes, with only one running
apache, the other 4 running squid...
Aside: I use Linux on quite a few boxes (p120, 16M) that I wouldn't dream of putting NT on, because it's such a hog. It would be nice to see something like this pointed out by these tests. -nme!
Microsoft would never have pushed for these tests if they weren't 100% sure they would win. They surely anticipated our response to sketchy testing procedures, and sure enough, we continue to lose. Yeah, yeah, this isn't a drag race, but it would be nice if we could program software that wasn't this much slower than Windows when one of the aspects of Linux we've defended is speed and stability. So, let's do it.
Calling out bogus battery capacity claims.
O.K., It seems that this guys are challenging the OSS comunity to develop new "futures" in the shiny code of linux. thats a fact, the guys of Redmond are jelous abaout the purity of code that the linux kernel have right now and how fast this could be changed and optimized. It is true that the IP threading stuff could and NEED to be a point of view to develop and optimize in linux, but, !!! hey Linux Kernel Coders !!!, what about to implement new code in the linux sources capable of optimize the performance of the kernel to succed an special kind of Benchmark test? I dont know, may be it could be an "extra" option on the kernel configuration menus, like the processor selector, but with options to optimize the kernel to a specific benchmark test. It could be a great idea.! Because all we know allready that the linux kernel right now has been demostrated a much more greater performance IN THE REAL LIFE than NT. So, if this guys are pushing us to "DEMOSTRATE" in the Benchmarking world who is the dominant OS, maybe we could develop a kernel specificaly complided to satisfy this stuff and kick an ass again...
Sorry about my broke english...
What would multi thread ip stack add besides serving thru 2 nics at the same time? Would it help routing? firewall? Other multi nic services in general? :)
Who can code it and when is it finished?
Exchange this!#$%#
but we already knew that. now we know where to make fixes. now we have a new focus. this is a good thing, because next time the tests come around, we still may not win, but we ill do better that is for sure.
after all these tests dont really mean anything, the NT machines are nothing like what they would be in the real world. they are so pimped out on MS engineered steriods that you will find machines like them no where else on Earth. but our happy little linux machines are everywhere, and they dont need a team of monkeys working around the clock to make them run.
i reccomend people reading something i read today but i cant find the link, anyways it was about how linux under the conditions in the worst mindcraft test could outperform its internet connection even if it had like 5 T1's hooked up to it. so basically even though they say NT outperforms Linux by X % it doesnt matter because the linux machine can already outperform its own connections... somebody post this url again if you know what im talking about it is good reading.
> ERROR: IEXPLORE caused an invalid page fault in module MSCONV97.DLL at 0137:01212d19. Stack dumped:
It's all a communist plot to overthrow the world.
-- My neighbors dog has a four inch clit.
The one comment that /. edited out of my original post of this story is that since BeOS allocates a thread to every network socket opened, I'm really curious as to how it will play out over the longer term.
I've used so many hackjob thread packages, that finding an OS with this technology engineered in from the ground up gives me a woody.
It is clear though that a lot of people don't appreciate the value that certain OS features can bring to the party. I attribute a lot of this to ignorance or religion. Little tutorials like yours on multithreading are invaluable in correcting this. Thanks.
It's like when I first read the source code to the Linux IP stack and noticed that most of the lessons learned by BSD over 10 years weren't visible ANYWHERE. No slow start, congestion control, or even OPTION HANDLING (it was all stubs). That has changed some, and will continue, but some people require things to be right before they'll commit.
I'm as prepared to like Linux as the next guy, but I don't let one OS' features (or lack of same) define the possible universe.
In the real world though this does become a problem. In a given web site some files are bound to be hit more often then other files and when you have multiple requests of the same partition you get into the same mess.
War is necrophilia.
As for the topic, I believe M$ knows-darn-well what Linux (and probably most every other OS and most major software packages) can and can't do, and actively exploit this info in any test it preforms against any competitors product. Open Source makes it that much easier. Bottomless pockets also help. That's what companies do to "compete".
I started out on an Apple ][ back in '79 or so. Linux brings me back to the Beagle Bros products that helped me explore more than what came shrink-wrapped. Linux calls to those who used to take things apart when they were kids (or still do) to figure out "how they worked". Alebit more fun back in the mechanical days, as VLSI and surface-mounted packages aren't that fun to look at (or pry off boards.)
Linux allows for people to really learn about computer technology at all levels without having to worry about licensing limitations, development software costs, or "The Software Police" (acronym escapes me.)
So where's the cost savings? Having technically adept and curious users that know how to explore their system to their want/need and adapt it to their own style of usage instead of expecting their OS to automagically give them what they need would help. I could get together with my friends and build a Beowulf cluster next weekend just for grins (and experience, of course) just because it can be done for exactly $0. Computer geeks do things like this as a hobby. Some get paid to do it. (sigh)
Plus ca change, plus c'est les memes choses.
Because MS/MC appears to have been so careful to research the underlying problem with that particular benchmark using that specific hardware configuration, they have illuminated a core scalability weakness in Linux. The lack of a multithreaded IP stack, according to their claims, slows even Zeus down to about the same speed as Apache. I'm convinced that they went into this *very* confident that NT would win.
Does anybody else suspect that the kernel gurus are, even now, working on clean, fast multithreaded IP code?
Geeky modern art T-shirts
OK. Folks, reality check. When we watch a football/basketball game, do we respect the guy on the losing team who whines about how the refs were unfair, or how they were decimated by injuries or otherwise makes excuses? No. We respect the guy who says, "They beat us fair & square, but we're gonna practice like crazy and take 'em down next year!".
NT beat Linux. Big woo. Did we really expect M$ to sit back on their laurels and let Linux eat into their profits? If you did, you're living in a dream world.
Let's look at the bright side, we're now officially in the big leagues. We're getting press coverage and M$ seems to be making it a priority to discredit Linux in favor of NT. However, that also means we have a responsibility to behave as big-leaguers. That means several things:
1. As of now, M$ is the enemy. THE ENEMY. Until now we've been able to sling barbs back & forth, but no one took it really seriously. It's pretty clear now that M$ is eyeing Linux the way they've eyed other would-be competitors. Since they can't buy & shelve Linux, they must resort to their only other tactic... destroy Linux. I guarantee you this is their goal. It's absolutely vital that we understand this & do everything in our power to stop it. That means....
2. We have to come together in a much more concerted manner to make Linux the most rock-solid, high-performance OS out there. If we get beat in a test, fine. We'll redouble our efforts until Linux thrashes NT in performance tests. We *must* do this or M$ will win the ultimate war.
3. We need to be as effective at evangelizing as M$. To a large degree we have been, but now that we have the spotlight on us, it's absolutely imperative that we come across as intelligent, determined, *non-whining* & mature. The management types who see us and M$-marketing must see that we are more than a bunch a snot-nosed, pimply faced rude, crass hackers if he is going to entrust his business to us. That means we need to start putting our money where our mouth is. No more "M$ sucks, end of story", no more profanity, just cold, hard facts of how Linux is superior *as a package*.
Since I'm a coder, I'm more than willing to do my part to contribute to the performance-enhancement initiative. Other coders should do likewise. Non-coders can help too. Educate yourselves. Help with evangelization & documentation.
As trite as this sounds, part of what made Linux great is also it's biggest weakness. Linux is the product of distributed, unmanaged, collective intelligence. While that has freed us from stifling burocracy & promises made by markting types, it also makes us vulnerable. M$ has it's guns pointed at us, if we want Linux to survive, we're going to have to be a little more organized.
We all can contribute something. We just need to mobilize (and have a mobilization center).
we had this story, a couple days ago: here... it has the exact same url. this is the same story. doh!
That is Bullshit - it's not a feature, it's nothing but worn technology.
I'd like to see how the tests round up when they benchmark NT5 with FAT32.
I am 19 and switched to linux right before my first year of college. A lot of my friends in college switched to linux because of that same reason. We have labs and labs of sparcs.
- "My name is Legion, for we are many" -Mark 5:9
I think the partitioning was used to allow multiple threads to access the separate allocation tables simultaneously rather than having to deal with any of the synchronization issues associated with having just one mammoth partition. This is fairly common in NT administration.
You're not really telling the clients they can only see a subset of the data, you're just telling the web server that it has to grab it's data from four different drive letters. Though with Linux this is a transparent feature, in NT, it's just something that you have to deal with.
From the PC Week article:
==========
Previously, we had found that Samba could outperform NT 4.0 using NT workstation clients (see "NOS crossroads," PC Week's Shoot-Out of network operating systems). As a direct result of that Shoot-Out, Microsoft's performance engineers were prompted to find ways to enhance file throughput from NT servers to NT workstation clients.
=========
Apparently, one of the reasons Linux hasn't been blowing NT out of the water as in previous tests is that Microsoft has figured out how to reconfigure NT to do better. I wonder if they haven't also made changes to the NT kernel or other underlying software? Of course, since we have no NT source code, we can't tell. OTOH, we *know* the Linux community is already hard at work changing its code, as is the Apache and Samba communities, and both the programs and their users will be better off in the long run. How long have people been doing Linux versus NT tests, and NT is just now catching up? How long does anyone think it will take for Linux/Apache/Samba to adapt and come back even stronger?
(Maybe Linux is more Borg-like than MS after all, but in a good way!)
Linux advocates definitely suffer from extreme myopia. As soon as anyone hears anything that slightly discredits Linux, it is dismissed as Lies, M$ funded, blah, blah, blah. I use Linux all day and avoid M$ products, but I still can come clean. No one will ever be happy with these tests until Linux wins and that is weak, face the facts.
its sad but that is how the business world acts.
:)
they see a threat to their market share and try to strangle it and kill that threat.
hence this hype about nt being faster. too bad nt is not efficient for the administrator nor is it reliable. moreover, the need to reboot a web server for every litle thing you change or add is ridiculous--for businesses this can mean downtime and downtime=lost revenue for some e-commerce sites.
NT=downtime
Sensei
Sensei
Linuxnewbie.org home of the NHF's
Regardless of the minor differences in speed, I still prefer using Linux over NT any day. At least I don't have to reboot when changing the smallest sodding configuration thing. Call me old-fashioned... but I like my command line. And I think lots of people feel the same. Show me an editor as good as vi native to NT :)
:)
All we need to do now is enhance the code slightly and beat NT. No worries
Uh, maybe there is the fact that you can get Linux for free, as many users as you want for free, and run lots of free program, without big brother looking over your shoulder. Anyone know what NT Server w/25 clients costs...like $3500 i think?
$3500 != free
i think not...viva linux!
JediLuke
-Do or Do Not, There is no Try
Here is an interesting question that really doesn't have much to do with the article but oh well.
When, as in what is the average age of the person, does someone switch over to Linux from MS?
From what I can tell there are a lot of people in the Linux community that are in college. My guess is that there is probably a disproportionate amount of Linux users to MS users in the college scene. I know that at the college that I go to, a great deal of the engineering students have gotten really ticked at MS and NT and there seems to be a very large number that are switching over to Linux.
One of the big reasons is that there are quite a few labs for upper division classes are run with SPARC stations, and with ethernet connections in the dorms students have found that it is easier to get Linux on their box. They just ssh over to the lab and work from their room. You can't get nearly seamless connection between two computers with NT, especially if the one that is the important one is running Solaris.
Disclamer - Opinion of Person
I hope some of you linux geeks are at least slightly humbled by this (I'm sure) unexpected article. I read the posts when the first test was run and 99% of the posts (I think) were saying that had the tests been run properly, Linux would have beaten MS hands down. Not true. Oh, but there were thread problems, blah..blah..blah.... Linux is great, I use it too, but I'm not so stuck on it that I can't find it's weaknesses, or MS's strengths. So what if Bill Gates is a Pig. That holds no real meaning when your company is choosing a product. And if thread-handling or whatever needs to be fixed before a linux webserver can beat an MS webserver, then, MS was ready for the test and linux wasn't. This seems to be the case up to this point. I wish that linux users could put the energy that they use to complain to better use, like fixing bugs.
PC Week says that "After a tortuous five days of tests, audited by the best and the brightest from Mindcraft, Microsoft and Red Hat Software Inc.,..."
So is Red Hat part of your ZD/M$ conspiracy? If not, then where is their denial/disclaimer? Is it just possible that they *did* audit the tests, and considered them to be fair?
(BTW, PC Week probably meant "torturous" testing, meaning painful and difficult, rather than "tortuous", meaning twisty.)
Someone should benchmark Linux and NT on some ancient hardware like a 486dx4 100 or something crazy like that!
On NT 4.0 you can kludge mounting multiple volumes under a single root using DFS (Distributed File System).
On Win 2000, you can just do it -- mount multiple volumes under a single root using so-called reparse points (basically symbolic links). Best of both worlds: you've got the drive letters as shortcuts, but you've also got the Unix-style hierarchy.
See http://www.pc-depot.com/neat/neat_971105_dfs.html
Many of you are missing the point entirely here. It is, however, quite subtle (the point). Before I share what I think it is, let me address posit my opinions on a couple of issues:
(A) Linux is just barely learning to do SMP. It was experimental less than 6 months ago. The fact that NT on a Quad XENON showed a higher degree of scalability is essentially meaningless. There are quite a lot of brilliant guys that will make sure that a year from now Linux SMP will show maturity and scalability. Don't sweat the fact that our baby can't beat the olympic runner yet. Let's at least wait 'till he is an adolescent to expect it.
(B) As another poster pointed out, Microsoft has teams of people that tweak the O/S purely for attaining high benchmark results. The configuration used in the test would NEVER be used in the real-world. The company has BILLIONS of dollars to throw away doing such nonsense. Don't sweat it. All our hackers were busy writing the next version of the kernel, not playing with benchmark hacks. It would have been much more interesting to see what the benchmarks are like for an Out-of-the-box NT 4.0 installation on a vanilla (average) PC against an Out-of-the-box Linux distribution on the same machine. My bet would be squarely on Linux.
Okay... Here is the point:
The fact that Microsoft made such a fuss in the first place means that our baby is in the big league. Microsoft is advertising for us! It's amazing.
As any good marketer will tell you, NEVER compare your product to your competitors (even if the comparison is beneficial) if your competitor is less well-known than you. Doing so gives your competitor a larger market presence.
In fact, if I were to express my paranoia (as so many of you have), I would say that Microsoft is using this ploy to attempt to assert that they have competition (having competition benefits them in the Justice Department Trial). My guess is that Microsoft laughs at us and thinks we represent absolutely ZERO threat, which is why Billy boy doesn't mind breaking the cardinal rule of advertising for your competitor. He doesn't think we really represent competition.
All in my humble opinion, of course.
-Dan
Petition reques
Ha ha ha...weeeeee!
F: G: H: I:, etc. Good point. However, you can make this transparent to the users with logons scripts and drive mappings, etc. It's not as sexy as mounting a single file system in Unix, but it works.
I'm currently logged in as my redundant backup account as my primary failed over.
Just out of interest, you can mount partitions into a folder in Windows 2000.
Use Windows 2000 before you make foolish statements.
They are so the same thing. IEXPLORE.EXE(60k) is there for applications which still expect internet explorer to be IEXPLORER.EXE. It's simply a wrapper for the EXPLORE.EXE file.
If you open up Explorer, and close the explorer band on the left, tada, you have internet explorer, with look and feel. If you type in an internet URL, tada you have internet exploerer.
Lets try it the other way round. Open up IE, goto view -> explorer bar -> folders, tada, the folders list appears in the band on the list and you have explorer. Or just type in a local folder address in the address toolbar and tada you have explorer.
THEY ARE THE SAME THING.
I would trust PC weeks tests any more than I trust
Bill Gates to make a quality product.
Most Pc WEEK/ZD Employees are almost incompetent,
just look at ZDNET for proof.Sorry this proves nothing to me .