Slashdot Mirror
Salon on Mindcraft II
Andrew Dvorak wrote to us with Andrew Leonard's latest Salon piece about the in-progress Mindcraft II testing. Interesting reading-and the difference between the OSes is evident-Linux sent hacks, MS marketers.
← Back to Stories (view on slashdot.org)
Maybe Microsoft can beat Linux+Apache+Samba in speed. But how can anyone even dare to use something from a company that can't even spell security (or bugfree)?
This is because M$ has found a testing domain
that at the current time favours the M$ setup.
1. Using a 4 processor machine.
Can you justify using a 4 processor Intel
machine, when you'd be much better off with
a cluster of 4 1 processor machines in terms
of reliability. Anyway, if you need to scale
that badly, get a Digital Alpha or something
like that.
2. Using a kernel level threaded webserver.
Would you want to have a process doing this
on your Linux box? I think not. Apache forked
processes die after serving a number of hits,
so if anything goes wrong the webserver will
return to normal operation after a few hits.
This is MUCH more reliable! Web serving is
about reliability in 99% of case, not speed.
3. Serving static pages.
Nobody serves that many static pages. If you
need to do that get a different web server
(Zeus?). The power of Apache is in things
like mod_perl and PHP3. Apache is more than
fast enough in what it does. You can fill a
T1 line with a Pentium 100, which is more
bandwidth than most sites have.
4. Choice of harware.
The hardware chosen has better drivers for
NT than for Linux. In the original test,
they used a beta driver for the RAID that
wasn't properly SMP aware.
5. Samba?
Samba is a cool product, but why should we
have to do windows file serving, instead of
our native networked filesystem protocols
like NFS? I'm not sure about this test, but
in the first one, they used Win95 clients.
Samba is NOT optimised for Win95 clients for
the simple reason that serious businesses that
can afford such a server would be a lot more
likely to be running NT workstation on the
clients.
6. etc...
Michiel Toneman
"Brown shrugged, and said a couple of programmers
had spent a weekend or two hacking together
solutions to the main problems demonstrated by the
Mindcraft test. The changes to the kernel weren't
yet stable enough to be employed in the current
PCWeek tests, said Brown..."
If the Kernel fixes are not used, aren't they
just waisting their time and verifying Mincrafts
previous tests?
Anyone see vegas odds posted on this one yet?
>The Linux people are not allowed to
>use any patches that did not exist (or were not >publicly available)
>before April 5, 1999.
April 15?, this must be when Mindcraft did their
test. They don't want to let any fixes that
directly address problems shown by the Mindcraft tests. Then they are guaranteed that the same results will follow.
Setup.
I cant get to the OpenBenchmark page.
Do they run MS IIS
These benchmarks are totally unfair. Apache is not particulary fast (because speed is mostly irrelevent in web serving) and SMB is native to Windows (even throw SAMBA is very good)
Let's look at the first impressions.
MS is saying that even a marketing guy can handle this.
Linux community is answering by sending their best hackers.
Boy, these guys are good at marketing. If you were an average IT/finance (more common than I thought) guy for a mid size company which one would you pick? (Remember the last good sysadmin you had?) But, times will change. It is wake up time baby.
Why focus on this microsoft controlled test, on hardware 90% of us can't afford?????
Why doesn't RH take some of there 100 Mil they'll get from their IPO and hire Mindcraft to do a reasonable test on reasonable hardware.
Basicly it sounds like MS went to their techs and asked them to describe a test where NT could beat linux, then hired mindcraft to perform that test.
One solution of course it to run Linux on a platform 99.9999% of linux developers can't afford, and are unlikely to optimize their code for. Guess who can afford $40,000 NT servers......
RH (or whoever) can play the same dirty pool a even use the same company and show that Linux trounces NT.
If you came out with results saying Linux on a $5000 server beats NT on a $10,000 server (for anything, even NFS file sharing (a more open standard?)), I am sure the media would pay attention.....
I cant get to the Slashdot site.
Do they run Linux?
No shit!
Isn't that precisely the test that you people were bitching about being unfair? I guess what mindcraft should have done was jump into the future and grab all the Linux patches to the flaws they found then perform the test.
I believe that a few years from now, when people look back at this first battle, and the attempt by MS to destroy Linux's credibility with a pre-emptive strike, historians will judge this as perhaps the greatest blunder in the history of OS wars. It will be compared to the invasion of Russia by Germany in WWII. By challenging Linux early, they hoped to knock the wind out of the Linux juggernaut. What they still haven't realized is that the open source paradigm is self-healing and adapts according to the challenges placed before it. To throw down the gauntlet before the masses have attempted to adapt Linux, allows the open source process to eliminate the shortcomings highlighted by the Mindcraft test. Linux will adapt just as the Borg. I guess we will have to reprint those old T-shirts with a penguin Borg ;-).
People use IIS on NT.
People use Apache on Linux.
More than they use Zeus.
Thus, Apache+Linux is a reasonable comparison to IIS+NT.
Actually wasn't it a publisher of Computer Games
around the heyday of the C-64? I remember some
titles. I think a book publisher like Simon & Shuster or something bought them out...
So if they look bad, whats to stop "holding" companies such as Intel and AMD use from buying
up tons of RH stock?
It's funny that when the Mindcraft tests first came out all the Linux zealots were screaming that they misconfigured the server, or whatever. But then when things were examined more closely and it was shown that out-of-the-box NT whomps Linux butt big time even when some Linux guru has tuned the machine to the max, the cracks in the Linux zealot camp started to show. Suddenly, all these other ideas for tests started to come forward -- but only after it was clear that Linux did miserably at this one. This sort of behavior is nothing more than a bunch of sports nuts standing on the sidelines and yelling "Go Team!".
I hope Red Hat was smart enough not to agree to a contest they can't win.
Why? Why so partisan? Why not hope Microsoft was smart enough not to agree to a contest that they can't win? Or maybe why not hope that both companies just want to get the truth out so that customers can decide for themselves which to use. Wouldn't that be best?
Mindcraft seems to think that is what it was doing in the first round of tests, before the Linux camp started saying all sorts of defamatory things about them. Now they want to clear their name. They did a second round to make sure that their results were sound, and when that wasn't "open" enough for the Linux crowd, invited them to show up directly. Personally, I think Mindcraft is being very gracious to the Linux camp.
Nonsense. Mindcraft isn't being "gracious". They took a much bigger PR hit than they expected and this is damage control. Mindcraft's first tests used hardware, tests, and tuning to favor Microsoft (their employer). Their second test used the =same= hardware and tests to prove that their tests were "accurate." Last I heard, they still insisted on the same hardware and tests. The only change in policy is to allow Linux experts to tune their system. Never mind that the tests have little relation to real-world and that the hardware was intentionally selected for poor Linux support.
Mindcraft wants to show that actual test results reflect what the OS can do GIVEN the hardware and test validity assumptions. Meanwhile, it is those assumptions that are fundamentally flawed.
Hey doofus -
If you can't get to slashdot, how did u post this?
Yeah, sure seems that way.
... 16? years ago, and then someone came along years later, rewrote it and then started getting all this worship and credit for pretty much the same thing you did so long ago, you'd be pissed too. I mean, what did you think he was gonna say "Linux is a great thing, its essentially the same thing I did a long time ago, but now its getting a lot of press and attention while my plan 9 project isn't. There is no point in me continuing my project because Linux is such a great thing that it will certainly be the OS of the future." Although I didn't expect him to use such a harsh sentence as "Microsoft is unreliable, Linux is worse."
But who cares?
If you can afford a 10k machine, sure go ahead and buy NT, not like you'll hurt anyone's feelings. Linux doesn't have to be the best at everything. Actually for server situations, FreeBSD seems to be a lot better although I haven't used it. cdrom.com and hotmail use it. Cdrom.com has the record for most data(clients?) transmitted simultaneously. Hotmail is probably one of the best examples of dynamic content out there. I don't think its ever gone down. Its owned by MS and they haven't changed the servers to NT... maybe a clue there?
I'm getting sick of all the "father of unix said Linux sucks" stuff. He said he looked at some of the code and saw it on some of his friend's computers. Maybe it didn't work well for him. Works fine for me, and that's all I care about.
Also quite possible that he didn't use NT machines all that much, and when he did, they worked fine.
Thing is, he didn't even say what was wrong with Linux specifically. Probably means he didn't do much testing or experimenting, comparing both OS's.
Anyway, for some reason, plan 9's website seems to be down. I probably missed a news article or something, but it seems like plan 9 isn't being developed anymore.
Not accusing Ken Thompson of anything, but if you invented something
I have only seen slashdot go down twice. Sometimes it gets a bit slow, but I don't see it go down often. Of course, it would be a neat experiment to see slashdot running on NT for a week or something.
That would be slashNoT?
Definitely. Sounds like it will sell.
Kurt Gray laments:
Even still we have no single big source of Linux performance help.
Have you checked out TuneLinux.com. It's making a good start at being a single big source of such help.
----
Open mind, insert foot.
At the time the Mindcraft paper came out, I was working for Linux Hardware Solutions. My first thought, upon looking at their numbers, was "man, that Dell SUCKS! I wonder how much they'd charge to benchmark one of our boxes against that Dell? We'll wipe Dell's rear!"
My next thought, however, after looking over the details of the "benchmark", was "man, I wouldn't want those incompetents near my computer." There are plenty of reputable benchmarking outfits. Giving Mindcraft more business is NOT a proper solution -- Mindcraft has proven that they are incapable of conducting themselves professionally. They poorly served Microsoft by doing a "benchmark" that had so many holes in it that even the press didn't believe it. Remember, their job was to generate numbers that would be believed -- and they failed miserably. Should we reward such failure with more money?!
-E
Send mail here if you want to reach me.
Posted by FascDot Killed My Previous Use:
"Any company that is setting up a server, whether it be a web server or whatever, and is willing to spend that much money is going to do some homework before they go out and buy something."
Unfortunately this is not the case. My company routinely buys $10k-$20k server hardware but blindly sticks NT on them. Furthermore, each of these monsters serves up one application.
We've also got a $10k NT machine that we are supposed to migrate to for fileserving (from Novell on a couple 486s). This project has been around for about 6 months and has gone nowhere since our network admin (a total incompetent in addition to our other problems) can't make it work. NT has no disk quotas, crashes the machine, doesn't support workable login scripts, etc, etc, etc. So what do we do? Don't use a real OS! Waste the money by putting the project on the backburner!
I actually setup a backup print server on Linux because the NT print server kept crashing so much (about 1/week). But the only people who use it are the ones I supervise. *sigh*
--
"Please remember that how you say something is often more important than what you say." - Rob Malda
but thats just because its a bit more relevant
to something i want to set up. anyway, im
surprised red hat even bothered if its
true that hardware is stacked in NTs favor.
i doubt MS would accept if the hardware had to
be agreed on by both sides. maybe the test
should go on for three months on a publically
accessable web server (the request is switched
off to both with identical content) with a few
equal CGIs (like a search). and count the server
crashes of course...
id love to see Bill Vs RMS on celebrity
death match! (or a real one for that matter)
I write alot of Client/Server Services. Mostly hitting Databases. I do agree that I love my work. Still I'm also a speed FREAK! Every Blue moon a Security guard will catch me running from PC to PC in the office. Runing what I like to call Load tests. With over 700 clients company wide I want my services to run as fast as possible. ( So they do not get replaced by MTS ). Even though the company does not care how fast it runs as long as no one notices a wait. I still like to do my own meger Benchmarks for myself. I think its very cool to come up with a tweak that will let me get 25-50 more tranactions in a minute.
I'm not sure why so many people seem to put down benchmarks. Benchmarks decide what CPU I buy for Quake, What video card I buy for quake, And what Database I use for a VERY Large size company.
Last one in jail is a fascist.
So three Linux geeks have been given a full week
to tune their machines. I hope they post everything
they do on the web so we can all pick up a few
performance tweaking tips from them.
I'm sorry but the "Linux Enterprise" web sites that
sprouted up during the first Mindcraft aftermath
have already petered out -- they posted a few
little tidbits and that's it. I submitted a full
page of Apache tuning tips to one of those sites
and they didn't even bother posting any of it.
Even still we have no single big source of Linux
performance help.
I hope TuneLinux.com takes off, but so far it
hasn't changed much since the week it was
launched. The x86 and TCP/IP pages still
only have two tune-up tips posted. The Apache
has three links to www.apache.org
What I'd really like to see is the Linux team at
PC Week disclose everything they did to optimize
their servers.
...but not about the benchmarks. Win or lose, the test will help Linux by shining a spotlight on problem areas so that they can be fixed. Microsoft will always be using the media to its advantage rather than make better products. It can afford to do that. Linux has to constantly improve. Tests like this will help it do just that.
It's not enough to bash in heads, you've got to bash in minds. - Captain Hammer
How many people use 4 100MB ethernet cards and serve only static pages?
Forget the antitrust trial. Ladies and gentlemen,we have a competition.
Perhaps this test and all the media hype around it have nothing to do with MicroSoft trying to squash Linux, but just to generate the illusion of competition.
Perhaps after the trial is over, we will see them take out their big guns.
How would one define a real world benchmark? -Specifically?
Some of the benchmarks are real-world benchmarks and some suspect. Or perhaps benchmarks are a pain in that they show one perspective on the "real-world". Is there any URL source with a long-winded analysis of the relative merits of different benchmarks?
Is there any effort underway by the open/free software community do publically develop a set of real-world benchmarks? If so, where can I find more information?
Life is like an egg better scrambled than fried. -- Ken Sawatari
Yes. Microsoft owns everything. They own Slashdot, too, and run it as a place for Linux hackers to vent their spleen ineffectually.
Or it could be that Salon is sympathetic to Linux and agrees with the contention that the test is set up in such a way as to give NT the most advantage possible.
Some of the Linux community is reminding me more and more of the old Amiga community, with the "every little thing they do is FUD" seige mentality. (You think a single Amiga 2000 couldn't redo all the special effects from Independence Day faster and at a fraction of the cost? You must be a real anti-Amiga bigot!) You see the same mentality in some Macintosh users, who cannot accept the possibility that one might buy a PC for reasons other than secret mind control messages embedded in "Intel Inside" television ads.
Not everything that everyone utters that fails to put Linux in the best of all possible lights is part of a grand conspiracy to spread fear, uncertainty and doubt around the universe. Sometimes the negative comments have ulterior motives behind them (which was the real argument against the Mindcraft test); more often than not they come from misinformation, misunderstandings or, at worst, willful ignorance. And there are cases, like it or not, where other operating systems do have an advantage for some users. Maybe you're not a member of the community that wants that advantage, and that's okay (the things that make the Amiga great for video production aren't important to you, or perfect color-matching between print and screen isn't enough to get you to buy a Macintosh). Maybe it's an advantage you would like to see in Linux, and you'll take advantage of the openness of the OS to add new functionality that narrows (or even closes) the gap. But denying the possibility that other operating systems might have some things that intelligent users could be attracted to--yes, even Windows NT, like it or not--is just generating FUD of your own.
We know that Micros~1 was not just targeting linux, but the linux/apache combination. They don't care about zeus since Zeus isn't causing their market share to drop through the floor.
As far as this four ethernet trick thing goes, I simply don't understand this. The linux equivalent would be to use gigbit ethernet, not four 100Mb cards.
What we should have is a price based benchmark. Give each side $15000 and see what numbers they can get. Include the costs of all the NT server + licenses as well. Testing on 200 clients mean they will have to buy a Emachine with 16MB for a server. Bye-bye, NT!
support gun control: take guns from cops
would'nt putting squid in front of apache drastically reduce the number of processes apache has to spawn for this type of test? Can't you do this on the same machine and see some performance gains?
support gun control: take guns from cops
I wonder who those (RH) people are. From the article, it sounds like they know their stuff, but obviously, people from the Samba and Apache teams would be the best.
Let us all wish them good luck now, should we? It's a little suspicious that MS/MindCraft was so interested in re-running the test. Perhaps they've got something nasty for us...
/* Steinar */
(This comment is of course GPLed.)
While we are all wondering what will the benchmark results be, we are forgetting one thing.
/Performance ratio
Let's suppose Windows NT gets a score of 100 and Linux gets a score of 95. Ok, M$ goes home with a happy face and prepares its press release. Linux geeks wonder what happen...
The issue here is, how much did the NT box cost and how much did the Linux one cost.
Thats what we call Bang for bucks-Price
Why would I pay a few thousand dollars to get an NT box that gets a score of 100 while I can get 2 or 3 Linux boxes at the same price.
.... end Jules impersination */
You seem to be missing a larger piece of the picture I painted, as that sentence wasn't really the crux of my post. But I'd like to take your points one by one.
I'm not sure how your seeing people talking about "how linux is free" has much to do with anything, but its still a fact. And if you don't think the bottom line matters, even at a company spending a 10 million a quarter on IT, you are sorely mistaken. Yes, Linux works better in many situations, however so does NT, and so does Solaris, and so does MacOS. The right tool for the right job. And in a, money is no object environment, the fact is that anyone would be hard pressed to come up with a soution better than Outlook 2000, Word 2000, Exchange 6, Windows 2000 Server. I've seen the demos, they are scarily cool.
And, as that was my example, lets expound on it a bit. The number crunchers where I work are currently looking at what it would cost to impliment just such a solution for the 900 users in the central office. Let me share with you the initial number they came up with to support each user for one year. $100 per month, per user. Still think it doesn't seem like a lot?
$12,000 doesn't seem like a ton of money for an enterprise email system to me either. But -$0- sounds like quite less. (Btw, 600 users doesn't sound like much of an enterprise either. Heck we have a truly stable Linux email solution right now supporting 10,000 users. Net cost initial: Software: -$0-, Hardware: -$50,000-, Setup: -$4,000- man hours. 1st year support: Est. 20 hours: -$1000-.)
I'm not sure when exactly all Unix (Linux?) administrators became 18 year olds. Personally I'm 28, and while I abhor day to day Unix administration its certainly what I got started on, and I still do some of it today. Did I get started on Linux? No. Can pretty much any real Unix administrator be a real Linux administrator in a matter of weeks, yes. And there are plenty of Unix admins left out there.
Do you have any idea what MCSEs are making these days?? Its obscene.
And finally. Linux doesn't need to be sold to companies It is selling itself day in and day out. Its stable, hence reliable, and generally getting easier and easier to use. Not to mention, free. Oh did I mention it was free? Did I mention the licensing for 99% of services was, oh, free? Linux doesn't need to sell itself. However, the applications must be written, they must be ported, they must make it over to Linux, and fast.
http://windows.scares.us
... but how you play the game.
Whether or not a Linux installation wins the benchmarking game against Microsoft NT, the fact is that Linux has made remarkable inroads in the past year - and shows no signs of slowing.
CEOs, Managers, CTOs, etc... are all making a decision to test Linux installations, if not outright deploy them. Electionic mail infrastructures are being built on Linux based machines in Fortune 500 companies. Winning benchmarks or not, the Linux juggernaut will roll on.
A free operating system, that runs free server daemons is going head to head with a company that would charge $12,000, just in licensing, to offer 600 users email services using thier groupware solution.
What would my message be to the Linux/OSS coders out there? Stop looking at benchmarks and start looking at functionality. I've seen the Office2000 suite. I've been to the Mircosoft presentations. If collaboration is the way of the future, then someone needs to come up with an answer to Windows 2000 Server, Exchange 6, and Office 2000 - and quickly.
All the benchmarks in the world won't mean jack shit if you don't have software that people want running on your OS to produce them.
http://windows.scares.us
Three markatroids and three hackers? Is that just for the press conference or is that it for the whole testing? Is MS going to bring in acutal engineers to do the tweeking. For that matter - are the Linux guys going to be able to call on the communities expertise? ;)) but going head-to-head - will be a steep learning curve. I mean, when the oppisition isn't a nebulous company but rather 3 guys and their server on the other side of the room - that's got to be a hell of a motivator. :)
Personally I think we're going to lose (well, unless the 3 suits really is all MS are throwing at it.
How much tweeking are we allowed to do? Easy to imagine a situation where the open source team start tweeking the source not just the config and MS are going to cry fowl. But for my money that's open sources strenght, it's not like there's anything stoping our hypothetical customer from doing that too.
Like I said - I think we're going to lose, but we're going to get alot out of it. I for one am eager to see what transpires.
AdamT
(As a side note - could we measure pages served per 24hrs instead of per second. Perhaps we can catch up during the reboots.
... with eskimo chains i tatto my brain all the way...
Why not fight FUD with Anticipation, Certainty and Knowledge?
How about tossing up some research-oriented web pages... oriented, that is, toward the middle management sort of people that are making OS implementation decisions.
Keep the site news-oriented, but keep the news oriented toward the latest developments in OSes and core support software.
Run a sidebar tracking some handy things... like the number of bugs found. Estimated cost of ownership (factoring in things like downtime and lost files). Risk factor. Etcetera.
Start running this as an alternative media site. The damned magazines aren't nearly aggressive enough in emphasizing the collosal and rather sickening security holes in IIs, and the safety/stability of Apache, for instance.
Call it ZedNet...
Anyway, point is this: once the actual cost in owning Microsoft software becomes apparent -- and not just the idiocy that they call IIs, but also the file effups of Word, the trashed files of Windows and the virus-luvin' Outlook -- then the corporate world will start to gain a clue.
--
Don't like it? Respond with words, not karma.
You are completely clueless, and do not know what FUD means. People like you, GiMP, give Linux advocates a bad name. It is astounding that you would present such an attack on a journalist who has demonstrated himself to be a friend of Linux. Don't you realize that it is people like you who are causing Linux to get poor press (by getting journalists pissed off at the whole Linux community)?
Get a clue: Linux is not currently the best at everything. Given the rapid rate of improvement, Linux will soon win in categories where it does not currently win, but if you are such a moron as to think that anyone who suggests that Linux is not currently the best OS for any conceivable application is in Microsoft's pay, then please just shut up.
There's a reason that the trade press just stopped writing about the Amiga: it's because any time they printed anything other than press releases from Commodore they were assaulted by people like you. So the path of least resistance was simply to not write about the Amiga at all.
The test is a replay of the Mindcraft test which is slanted to favor NT. The guy who did Samba has already said that, for this case, NT beats Samba. It is not FUD to print the truth.
While Microsoft has a near-monopoly on the desktop, they are not doing nearly so well in the server space, which is what tests like these are about. Linux or Solaris boxes running Samba are direct competition for NT Server seats, and are winning that competition in many cases.
The existence of competition in some of Microsoft's markets doesn't help them in claiming that they don't have a monopoly in other markets.
WTF does this article keep saying that Windows will beat Linux, the tests aren't even completed.. God I hate FUD, is Salon under MS's claw too?
Everyone seems to be taking this for granted - but didn't the post-Mindcraft I analysis focus on some rather obvious blunders made by the company?
Have those blunders mysteriously vanished? Or do we now think they don't really matter?
D
----
Even if Linux doesn't beat NT, if it comes close, then it has won. The reason is in the cost.
This comment is indicative of how the Linux advocacy party line has been affected by the Mindcraft tests. Before the tests it was conventional wisdom that Linux/Apache would pretty much stomp on NT/IIS. Now it's some cost/benifit ratio calculation.
Bottom line is who cares? I would guess that less than 1% of MCSEs are even aware of this study. Performance benchmarks are only going to make a difference in a small number of cases. If I was making the Linux/Apache versus NT/IIS decision, there would be a large number of other factors to think about first, such as what your developers know for example (ASP/VB or PHP/CGI), or how the reliablity or managiblity stands up.
The fact that some in the Linux community have gotten into a pissing match with MS on this one only shows that they're playing the game by Microsoft rules.
--
Business. Numbers. Money. People. Computer World.
I thought Mindcraft II was a computer game?? LOL!
This kind of competition is actually a good thing for those of us more interested in using computers than arguing about them.
On the OSS side, it means intensive research of a kind that will expose flaws and bugs that are not so easily discovered in everyday use - which means they'll get fixed, as we've already seen from MindCraft I. Meanwhile, the peer review process of OSS will continue in parallel, of course...
On the Microsoft side, it means that bugs and flaws will be exposed to the public eye, meaning they will have to fix them to salvage their reputation in the public eye...
I don't think that it's particularly good for either the Linux-companies or Microsoft's profit lines, but we'll see what happens. I think chances are it will cost research $$$ on both sides and not really affect sales of either side very much.
But I'm not trying to sell Linux, just use it - and sometimes work requires me to use Windows... so for me, this is a good thing. I say bring on more competitions like it!
--Parity
'Card carrying' member of the EFF.
I am not so sure why salon keeps hinting that Microsoft will be seen in a "positive light" though...
The whole tone of the article implied that NT was going to come out ahead. After initial remarks about the deck being stacked, the whole remainder was along the lines of 'NT is going to win this round... but Linux will prevail in the long run.'
Although it wasn't written that way, there was also the slight implication that given its inevitable defeat this time round, the linux guys had to get their excuses in in advance.
don't be too disappointed, the guy who did the tests showed up on linux-kernel and asked for advice.
It seemed clear (I saw some of the results, but I don't remember) that the mindcraft results are not completely wrong. OTOH he also didn't tune linux in every possible way and was told that, but then he didn't have the machines anymore.
what type of machine does Rob run for slashdot? With the traffic it gets, I'm just curious.
According to http://slashdot.org/articles/99/ 03/08/170243.shtml, the server is a VA Research Dual PII/450 with 512 MB RAM. IIRC, the most common cause of the slowness is a lack of bandwidth, and the cause of the downtime is the ISP.
Mike
--
Mike
--
"Wi nøt trei a høliday in Sweden this yër?"
Is it going to compare the same server on two different OSes? The best each has to offer? Static HTML? Dynamic? Win9X clients only? A mixture of services and clients? High-end hardware? Hardware supported by both OSes? I hope Red Hat was smart enough not to agree to a contest they can't win.
Best we find out the details and register our complaints before the results are out.
Sheesh, evil *and* a jerk. -- Jade
... at this comment:
"Microsoft customers had already demonstrated
a distinct lack of interest in having a
version of Microsoft Office that worked on
Linux, so Microsoft had no plans to push
forward in that market."
Well, duh! Microsoft customers, last time I
checked, used WINDOWS. Hello!
--- witty signature
I don't understand why they even care. Let Microsoft conduct their benchmarks, just continue delivering the product. Everyone knows that Linus and RedHat has been doing that before, and think of all the FUD they have been going through. Microsoft tactics are not going to change. There are always going to be some stupid Win-benchmark where NT is secure, Office is fast, and IE is Sun-Java compliant. Linux doubled and doubled the number of its users for years while the trolls in the media always said how much faster their NT installation went.
no flames,
mxk
I was thinking of how to intentionally fail my drug test... It would make a good memoir story someday.
What's the worst that can happen? Microsoft wins. In a fair test, Linux should come in very close. And Microsoft gets to trumpet that "Windows NT is 1.2 times faster than Linux!"
But think about it further. If Microsoft wins, the hackers will tighten Linux up even more. Linux is leaner and faster than anything Microsoft will ever hope to put out.
And finally, Microsoft can't win in the long run because Linux can't lose! There is no company to go out of business and take Linux with it. People will always hack Linux, because they can. The kernel will always be there, and the GNU tools will always be there. Red Hat, SuSE, Debian, Caldera, and all the others could all go out of business tomorrow -- and you'd still be able to download Linux, compile it yourself, and put together your own distribution.
In short, Linux will never die.
Pretend there is some witty statement here.
>2. Using a kernel level threaded webserver.
The Linux guys might already be doing this.
There IS a threaded kernel httpd. Both RedHat
and Microsoft know it. The question is: are the
linux guys allowed to use it?
( http://www.fenrus.demon.nl )
I read a lengthy explanation on Mindcrafts website .. the "rematch" will actually consist of (IIRC) three seperate sets of tests.
In the first set, they would try to reproduce the results of the initial set of tests.
The second involved the Linux experts tweaking the Linux box to its max using any fixes/patches etc available at the time of the original tests.
The third involved the Linux experts tweaking the Linux box to its max using anything and everything available right now.
I dont have the URL, unfortunately; its on my (Linux) home machine.
I think Salon misunderstood Brown's "arrogance" -- Brown didn't really care. Benchmarks, etc. are interesting but irrelevant. Speaking as an open source contributor, I program stuff because it's fun, because I get a "kick out of it," and because I'm not satisfied with the current status of software. We're making this stuff for us, not to bring down Microsoft or win converts. Really, as long as there were a solid base of open source programmers, I personally wouldn't care if every other person and corporation on earth ran Microsoft software.
It's by us for us, the rest of the world be damned. We geeks think it's cool.
Regards,
QBobWatson
Joe Rabinoff
Retail bookseller Barnes And Noble has confirmed reports that they have a "task force" designed to counter the competition of local libraries.
Of course I made that up (please don't sue me). But it provides an analogy for "Microsoft vs. Linux". Although for-profit players like Red Hat, Penguin, VA, etc. are certainly involved now, the software collectively called "Linux" is still free. Most of the contributions have come from people who give away their work for the good of the community.
That phrase -- "the good of the community" -- is important. I liken Free Software to a charity project. To treat it like "competition" is like a cafeteria trying to crush the local Salvation Army soup kitchen. It exists because people want it, need it, and are willing to contribute to it. It's a community, not a company.
I know that this comparison is not pure. Many Free Software advocates openly express hatred of Microsoft. Many openly state that they would love to see their software bury Microsoft. Some of this may be due to "hate-the-rich-guy" syndrome. Some of it may be due to very real philosophical and technical objections. I wouldn't expect any business to take this lying down. I would expect a defense -- but I don't think "defense" is something Mircosoft plays well. Instead, they counter by going on offense against Free Software.
Still, to me at least, it's like watching Sears tell people not to get clothes from the local Goodwill. If some people like the Goodwill clothes better, whose fault is that?
Save the whales. Feed the hungry. Free the mallocs.
Oh yeah, believe it now . . . they are making a version for linux . . . I would not be surprised if it is 'redhat' . . .
It is often easy to tell what ppl are thinking by what they deny. The very fact that they mention it means they are at least thinking about it.
meant to imply something like 'redhat certified' or thereabouts.
not a bitch about rh at all, just the perception that it is the be all and end all of linux stuff, which is not of their making really.
I think m$ owns a bit of sco as well (and maybe sco is xenix?) so, if rh ipo's, m$ could easily buy stock . . . they have the cash and the history.
Maybe m$ on linux would be a good thing in the end. IBM seems to be able to compete across platforms these days and has seemingly given up trying to own the OS.
The thing here is not so much if Linux will kick NT (in this config, I doubt it will), but if Linux will do substantially better than it did in the first test.
:) kick NT's butt in this kind of benchmark(et)ing...the deal with the new test is...did Mindcraft f'k up the first and second tests? The answer is, almost certainly, yes...and that is what the community should focus on. NT may be faster this week, but Mindcraft has been lieing for long enough.
If the difference between NT and Linux is NOT the 3.5 times and 2.5 times faster that they published before, all criticisms agains Mindcraft still stand.
We all know that Linux will, sooner or later (probably sooner
Vox
Pain is the gift of the gods, and I'm the one they chose as their messanger...
If I remember correctly, NT Server outperformed Linux by a significant amount--more than 2 times (???). It would take some serious optimization to create a performance increase of that magnitude, especially since, I'm sure, M$ is doing more optimization of their own.
I don't doubt that NT would outperform Linux on such a beast of a machine, but I certainly wouldn't run NT or Linux on it (4 processors, 2GB RAM, a RAID). Isn't the SMP in Linux very new? I'd probably, well first shop and compare, then choose something like Solaris or another web server and OS optimized for that many processors, which would outperform both. Linux and NT, I would think, would be best at doing general purpose computing.
Although this server is designed for NT
This appears to be the crux of the matter,
though it's not just the hardware it's also
the nature of the tests themselves.
I wonder if they would consider getting the box
to do web, file and print serving at the same
time
(There are also the fiddles the original tests
involving memory and network cards to consider.)
Even if Linux doesn't beat NT, if it comes close,
then it has won. The reason is in the cost. How
much does NT cost compared to Linux
Or even "how big a Linux *system*, would the
money needed to but this quad xeon, NT (version
NCC 1701) and all the client access licences,
buy you?"
Rob wouldn't be able to afford NT to do all that he does.
It seems that NT runs better on some servers, but it better run ALOT better if I'm going to pay the expensive amount to have it.
Side note: what type of machine does Rob run for slashdot? With the traffic it gets, I'm just curious.
Steven Rostedt
-- Nevermind
There's no question that the average Joe Schmoe computer will run Linux faster than NT. Unfortunately, this "bench-marketing" is done on a high end server. Although this server is designed for NT, it would be good for Linux to do well.
:) And I don't need to mention anything about security.
Even if Linux doesn't beat NT, if it comes close, then it has won. The reason is in the cost. How much does NT cost compared to Linux. Also, you get the source (although Microsoft said they would give you the source too, but for how much???
If I was a manager, NT had better cripple Linux, otherwise it is not worth the money. So if Linux can keep up with NT, then it has won. It will only be a matter of time when Linux will equal the performance or surpass it altogether.
Steven Rostedt
-- Nevermind
agreed. If they insist on using Apache, they must be taken to task for claiming "NT vs Linux web performance". Clearly they are measuring Apache, not Linux, and in an area upon which it is not focused.
That two obvious solutions to this problem exist points out a rather blatant lack of objectivity on the part of Mindcraft:
1) Use the fastest available static page server for Linux, as is done for Micros~1. (compare both OS's under their best conditions)
2) Use Apache on both platforms to eliminate the rather massive application variable. (compare the two OS's in an "apples to apples" test)
Unless, of course, the whole point of this test is to provide the documentation for a misleading MS ad campaign that claims NT is a faster web server platform than Linux.
Geeky modern art T-shirts
I have been running Winslow on a few different PC:s and it has allways been a great performance hit to upgrade them to Linux.
No benchmark can change my first hand personal experiences about how Linux performs.
//Gunnar
--- Linux or FreeBSD, it's like blondes or brunettes. I like both. ---
Seriously, I would buy that t-shirt.
Wouldn't you?
Will in Seattle
Will in Seattle
I don't worry about Linux's survival. I worry about its acceptance.
I like linux. I use it at home. I use it at work (albeit a 'rogue' installation). I would like to see more support for it.
When a vendor comes up with a solution that my organization wants to use... I would like to see it available for my linux machine. When new hardware comes out that I'd like to use... I would like to see support for it in linux.
When my boss wants to implement some solution... I would like him to consider Linux as a candidate. If Linux won't fit the bill, so be it. But it would be nice to see decissions made beyond a single company's glossy brochures.
Linux "winnning" in the marketplace helps all this happen. Heck... for that matter, what's good for Linux is also good for Unix in general. I've heard less people use Unix in a "yet another dead OS" tone since Linux has made buzzword status in the press.
So no... Linux won't die just because MS frowns in its direction. That idea is absurd. But MS can prevent Linux from expanding in amazing new directions. The only harm is the opportunity lost. And what a shame that would be.
The thing is that they use the fastest server for NT (IIS) and we aren't allowed to use the fastest for Linux (Zeus?).
And then they use these results to prove that NT is faster than Linux...
Apache wasn't built for speed, rather flexibility, so it's pretty pointless to test Linux vs NT with apache.
My opinions may have changed, but not the fact that I am right
I think its a great match up to see a bunch of young hackers in the same room (and on the same level) as these tired old ms execs, and their hired help...I am not so sure why salon keeps hinting that Microsoft will be seen in a "positive light" though...
Correct me if I am wrong, but isn't the test server a multiprocessor x86 machine that costs >$10000?
Frankly, I don't see why everyone is so worked up over a test like this. Any company that is setting up a server, whether it be a web server or whatever, and is willing to spend that much money is going to do some homework before they go out and buy something. Anyone in the computer industry knows that with that kind of money more powerful machines start to really enter into the field, like Alphas, Ultras, and SGI workstations. All of which, when loaded with Linux or their respective OS, will beat out a x86 machine loaded with MS any day of the week.
As for companies that aren't as computer savy, Linux's older cousins like Sun, SGI, and SCO have extensive marketing teams who's job is to show the customer what is out there at that price.
If MS wants to conduct feel good tests in order to improve moral in their company, fine, I could care less.
As for porting Office 2000 onto Linux. I hope they don't port that piece over to Linux. Even if they do, who is going to spend money, year after year, on junk like that when Corel is giving out a better word processor for free?
Disclamer - Opinion of Person
I agree with you. Microsoft has virtually unlimited resources to spin the media any way they choose. Just look at all the commercials on TV showing how normal people are better and better daily because of Microsoft. The Linux community does not show commericals like how Linux has accellerated open source, given new life to older systems, drastically reduced IT costs(MS assertion that a free OS is negligable to total costs is ludicrous, considering NT per seat costs), and actually has given a choice in PC operating systems.
This Mindcraft fiasco is nothing but good for Linux. I agree with Salon in its assertion that a negative outcome would double the efforts of Linux coders and advocates. If we(Linux) win, that will be great but MS could easily spin the stunt or make another benchmark. If the latter is the case, I think MS will be smarter than to show how poorly they set this up.
-Clump
I can't believe I didn't think of that. Who using Linux would be a Microsoft customer? Someone who downloaded their beta media player? Here is an honest question: How can a Linux-only user be a Microsoft customer?
-Clump
The German computer magazine c't has also
:)
announced a "highend webserver test".
The nowhere mention mindcraft, but the description
of the test is:
4 CPUs, 2 GB RAM and RAID system
NT with IIS vs Linux with Apache
The magazine is supposed to be out at 21st
of June.
The c't magazine is a very reputable and competent
source of information, I'm very curious how
their tests will turn out.
I'm also confident that Linux will look good in
a fair test.
-Stefan