Stepping up its campaign against the Linux operating system, Microsoft has released file and Web serving benchmark results that, it claims, show that Linux on the mainframe lags behind Windows 2003 on Intel systems in terms of performance for the money. Read the report at ARNnet.
Microsoft Benchmarks Step Up Linux Assault
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
not that it isn’t possible that windows outperforms linux in this area
but I’d never trust benchmarks coming from microsoft
The mere fact they are going out of there way to prove that Windows is faster proves they are threatened and scared. The only problem here is that any data coming from them is a lie. I know, I worked for the devil a few years ago. The truth in their world lies between the scum and the money.
Anybody remember the lies they spewed about 5 9s uptime?
The linux community ripped through those in no time.
M$ should release the exact specs and instructions for a non-biased 3rd party to test ASAP. We wouldn’t want customers trusting this without outside testing.
…and compare servers running in (or out) of kernelspace on both systems. That sounds like a fairer comparison to me.
no posted methodology. basically, my product is better than your product, so there! no real news here. just kiddies on the playground.
i see it now my bad ^^. too bad the article did a bad job of summerizing it.
May 80386 at 33 MHz is faster than Xeon at 1.2 GHz!!! Believe me although I didn’t provide any evident because I say so and the benchmark was carried out by a friend that I paid.
If I’m considered trolling how about that news itself?
Let’s wait until kernel 2.6 comes out…
…and compare servers running in (or out) of kernelspace on both systems. That sounds like a fairer comparison to me.
Why? 2.6 isn’t even out yet (not ‘final’ anyway). That’s like saying it would be unfair to compare the latest version of Mozilla to IE6, since IE6 was released more than two years ago.
a Wintel vs Lintel comparison in performance? I think it has already
been done but I am not sure where. It would have to be from
a *relieable* third party though.
Well, MS seems to be making a big thing out of this…but if their “edge” (if there is one) only lasts a couple of weeks (2.6 is supposed to be out before year’s end, IIRC), then it’s not a wise thing for IT managers to make a decision based on such an edge…expecially since the next version of Windows is quite a few months (if not years) away.
BTW, unless I’m mistaken, their Web server runs into kernel space. Did they test it against a Linux kernel space web server? Otherwise, it’s comparing apples and oranges.
Are we are suppose to believe this tripe. As someone said above my post. They are very scared and threatened to even try this tactic to prove something. They are really worried.
Turn off the lights the party will soon be over…
Why would you compare Linux on mainframes to Wintel in terms of performance for the money? Obviously mainframes are ridiculously expensive, but you buy them for stability and lower maintenance costs. Compare Lintel to Wintel or STFU.
This is like saying a Hyundai has better bang for the buck than a Porsche.
Mainframes are not good at this type of workload.
Do people really trust what M$ says?? Bill just wants $$$$$ from our wallets.
Now MS is benchmarking Windows vs Linux, haha
Are they scared or what?
Is Windows 2003 available? If not, and if it is the upcoming product from Microsoft, then why not test the upcoming product from opensource developers(in this case meaning Linux kernel)?
I’d like to see Windows Server 2k3, Vs. Linux Kernel 2.6, Vs. Apple Mac OS X.3 on a duel G5.
Maybe i should have read the earlier article about releasing Windows Server 2003. Sorry.
…and compare servers running in (or out) of kernelspace on both systems. That sounds like a fairer comparison to me.
AFAIK, Apache can’t run in the kernel space and the small httpd that was included with the kernel was taken out of 2.6. Anyway, I’ve tried a 2.6 test kernel, and it still suffer from a lack of maturity.
God how i hate you linux zealots who don’t know jack and don’t want to believe anything that comes out of anyone other than linux supporters. Before any of you start spewing out YOUR lies about how these benchmarks are lies, get some proof. Linux by nature is SLOW. Slap on some graphical components into its kernel and it’ll be a dog. It barely wins in certain benchmarks is because it doesn’t have to deal with graphics.
Linux vs windows benchmarks are hard to find. If linux was a clear victor in these benchmarks we all know there would be at least 10 Microsoft hating linux websites publishing benchmarks on a regular basis AMD vs Intel style. It took time but I have dug up some of these benchmarks and found that windows is very strong in server benchmarks. These benchmarks are always met with like a thousand penguinists spouting poor optimization and death threats but none of them ever seem to just shut up and deliver. You think Linux is faster? Fine post your much less biast benchmarks here.
Put up or shut up folks.
Windows 2003 outperforms Linux on the mainframe ? Fine, now let’s go back to information that matters : What does http://www.microsoft.com run ? Linux, hosted by Akamai. Enough said 🙂
.. the emperor’s new cloths.
Well, do you have proofs that Linux is SO slow? I don’t want to start an argument, but if you say they should find proofs, well, perhaps you should too.
in what way?
X can be.
Linux is not X.
Slap on some graphical components
Why does a webserver need a GUI?
into its kernel and it’ll be a dog.
Why would you want to stick a potentially unstable, and unrequired item in the kernal?
you start spewing out YOUR lies … wins in certain benchmarks
So not lies then, but good choice of what to use its CPU time on.
Strange mouse
I’d trust this as much as I trust IBMs “my is bigger” specs and benchmarks.
That said IBM reponce is good, (paraphrased) “they are answering a claim we never made, and using kit in a way we would not” Funky.
Here’s a quopte for you
However, Microsoft was unable to reveal at least one important fact relating to today’s numbers: the source of IBM’s alleged claim that a z900 processor running Linux could perform as well as three or four Intel processors running Windows.
They put words into IBM’s mouth and then debunk that claim with a research that was produced by a non-independent firm. Furthermore, the research is nowhere to be found (z900 on Forrester site gave no results). Even if I wanted, I cannot believe this kind of information.
Wierd, people who use Macs actually believe Apple’s benchies though…
However, Microsoft was unable to reveal at least one important fact relating to today’s numbers: the source of IBM’s alleged claim that a z900 processor running Linux could perform as well as three or four Intel processors running Windows.
Funny they compare Linux running on a VM on a 900z when what they should have done was compared Linux running natively on a s390 or would that just discredit their claims altogether?
Ok, if Microsoft is going to play that way, why don’t I compare Windows 2000 running on Linux (using VMWare) vs. AIX on a Power4 system?
Arn’t Apples own benchies pritty bad too, and they (apple) admit they are using a less than optimal configuration for Intel systems (using GCC, not Intels compiler).
Arn’t Apples own benchies pritty bad too, and they (apple) admit they are using a less than optimal configuration for Intel systems (using GCC, not Intels compiler).
Well, I don’t want to be the Devil’s advocate (especially because I’m a PC user
, but… Who uses ICC? AFAIK, only scientific and benchmark applications are using it. At least GCC is used in the real life.
RE: Wrawrat (IP: —.130-201-24.mtl.mc.videotron.ca) – Posted on 2003-09-05 05:46:55
[/i]Well, I don’t want to be the Devil’s advocate (especially because I’m a PC user
, but… Who uses ICC? AFAIK, only scientific and benchmark applications are using it. At least GCC is used in the real life.[/i]
Most software companies I know either use Microsofts compiler or Borland. The reason why they compared using GCC is so that people can’t say, “oh, so and so used xyz optimisations” or “zya compiler is more tuned for that platform”. What *SHOULD* happen in theory was the GCC compiler *BEAT* the PowerPC as the PowerPC port is not as mature as the x86 one, however, what this does prove is even with a partially optimised compiler, one still will expect a decent level of performance.
Maybe when things settle down we will see a ICC vs. PowerPC Compiler (from IBM) comparision.
I downloaded the pdf which is given as reference. Turns out it compares Windows NT4/2000/2003 servers. Win2003 is massively faster (often 5-times) than its predecessors in the performed tests. Linux on the other hand is not even mentioned in this report.
What *SHOULD* happen in theory was the GCC compiler *BEAT* the PowerPC as the PowerPC port is not as mature as the x86 one, however, what this does prove is even with a partially optimised compiler, one still will expect a decent level of performance.
Maybe when things settle down we will see a ICC vs. PowerPC Compiler (from IBM) comparision.
I completely agree with you. I hope the PowerPC compiler won’t be used exclusively for benchmarks, though.
I completely agree with you. I hope the PowerPC compiler won’t be used exclusively for benchmarks, though.
What would be also good for Apple would be them to continue giving away xcode and GCC, however, for a small fee, say, $200, sell an xcode “professional” version which would include the IBM PowerPC compilers plus some extra xcode add-ons.
For obj-c, since it is a derivative to C, creating a compiler spawned off the IBM PowerPC C compiler base should be a rather simple operation.
Just as it didn’t matter whether NT was faster and more stable than NetWare, or it wasn’t. Whether it had better features, better user and filesystem management, more practical utilities, or not. All these didn’t ever matter, and even though in all areas NT was lacking, it outsold NetWare. Because it was hyped through the roof.
And similarly, these benchmarks don’t matter, regardless if they are true or not (my money is on “not”, but that’s irrelevant, as I said).
in this case the hyped one is Linux
God how i hate you linux zealots who don’t know jack and don’t want to believe anything that comes out of anyone other than linux supporters.
I view these discussions as debates. In this case you have the Linux vs. MS crowd. In other debates you have Apple vs. MS, or BeOS vs. Meaningful OSes. Everyone has an opinion and everyone wants the other side to just shut their flap holes and go away. That’s all fair I suppose.
I would suggest, to keep things friendly, that you not make broad statements such as “Linux zealots who don’t know jack” and then turn around and prove yourself to have that same level of “jack-less” experience. It doesn’t score any points for your team. Also, since you are every bit the zealot with your opinions as Linux users are with theirs, I would stop the name calling too. It makes us look no better than politicians.
Before any of you start spewing out YOUR lies about how these benchmarks are lies, get some proof. Linux by nature is SLOW.
The only benchmark in the world that I trust is the me-as-the-end-user-performance-benchmark. Using the MATEUPB benchmark, I have discovered that Linux and the BSDs are responsive and I have found that Windows (even the proclaimed panacea of all operating systems, Windows 2003) are sluggish. I would like to prove this to you too, but I can’t. You see, the only comparable MATEUPB benchmark available to the rest of you OSNews readers is the YATEUPB benchmark, and it may show different results. To each his or her own.
Slap on some graphical components into its kernel and it’ll be a dog.
So, that’s what makes Windows so sluggish.
It barely wins in certain benchmarks is because it doesn’t have to deal with graphics.
For me, Linux and BSD win the MATEUPB benchmark, and that is the only one that matters.
Microsoft has the habit of promising that a product will soon be released, only to announce later that it will be delayed (Longhorn anyone ?). Some call it vaporware.
Now they’ve perfected another craft : paying for a report testing their own software (Windows 2003 vs Windows 2000 vs NT 4) and then using it as a basis to belittle Linux. I believe it should be dubbed “vaportest”.
I’ve read the pdf file (60 pages long, Jesus !) and as indicated by Guido Schimmels, Tux is nowhere to be seen in it. Maybe Robert McMillan applied cut & paste to the wrong bits of information, maybe it was simply another instance of Microsoft bad mouthing the competition (IBM and Linux).
Anyway, once again, when we’re in presence of crack induced delusions, Netcraft is the methadone of choice.
MS Windows would have to be 10,000 times faster than Linux before I would consider paying one red cent. If it was only 9,999 times faster, than they can kiss off, the boggarts (HP).
There are two different “researches” mentioned in the article. The one comparing different versions of Windows is by Veritest and a link to it is available in the article. The “Windows beats Linux” one is attributed to Forrester Research and it is nowhere to be seen.
Would you please describe me, why do a server need any GUI/graphical stuffs and mediaplayer???? Sit back and play with your solitare, and don’t flame on the big boys site. 😉
M$ is the reason there are so many desktops around the world. Had we waited for linux…. then i bet there would never have been an IT boom or this hype of internet.
M$ should get a pat on its back for this….
and abt this benchmark???…. do they think we are fools here?…so lets see what they have got to offer.
Patience guys… patience…
>>M$ is the reason there are so many desktops around the world. <<
Wrong – If Microsoft weren’t around (and sprecifically their business practices) we probably would have more diversity and choice and real competition – so there could have been more of the AMIGA OS, Apple, Atari Gem (very good by the way i know musicians who still use their old atari)and not to forget OS/2 (which by the way was far superior at the time to Windows). So please just because Microsoft dominates (as in controls and tyranny) the desktop market does not mean that we would have not desktops without MS, that is plain rubbish.
>>then i bet there would never have been an IT boom or this hype of internet.<<
so you also imply that MS ‘invented’ or at least was the driving force behind the internet – again you could be not more wrong – fact is that MS almost ignored the internet not even having their own browser – till they realized that ther was another opportunity (long after others have, f.e netscape) and the rest is history (again using their capital and influence to either buy out or crush existing comeptition.
Everybody knows that:
– IIS runs in threads instead of processes
– IIS runs partly in kernel
– Apache runs entirely in ‘user space’ and is forking
process for every request
Nobody, with even a tiny piece of a brain, would like to run web server through OS kernel, it is dangerous in many ways.
Performance, security, stability.
Forking process for every request has benefits, for example, file system locking is not posible with threads.
The same things goes for Samba. Samba is a user space process too, while microsoft network runs in kernel. They should compare NFS on both platforms.
>>This is like saying a Hyundai has better bang for the buck than a Porsche. <<
Actually, I think it’s more like comparing a Porshe with 18-wheeler. PCs are very fast for individual users, but for heavy loads, you still need a mainframe.
Why not compare apples to apples… Linux on Intel, and Windows 2003 Server on Intel?
If they really are making a case based on Mainframe VM vs Win2003 or perhaps native OS on Mainframe vs their OS on Intel they *might* have a point.
I worked for a long time on Midrange stuff (IBM AS/400) and if they are saying (and perhaps when they say this one should sepearate out this as being an attack on linux, from merely attacking a prospective market) that they have a better solution, not only do I think thats ok, I think that they may be right.
Desktop and server side Intel based hardware moves at high speed. Its advancing at a great rate and why should a company not come along and say ‘hey, our small box is cheap and blows your mainframe away’.
Its been happening before microsoft existed. Most mainframes if they have any age on them are probably indeed beaten by these options, and today I guess you would have to be careful about what machines you buy for what purpose. A few years ago a mainframe might have been your only option, I doubt that remains the same situation today.
AdmV
It’s all about marketing. And who ever makes the best marketing pitch convinces the idiot CIO to buy their version of binary spew.
Is there even a comparison to be made here? Who in their right mind would compare the jobs PCs are good at with the jobs that Mainframes are good at? Windows on PCs will never have the rediculous sustained throughput of Mainframes, nor will Windows ever have the reliability of Mainframes. Microsoft’s customers simply do not ask for the same things that IBM’s customers do (i.e., Microsoft’s customers don’t care about quality–they want it cheap and now, like a used Chevy).
Hmm. Winders on Intel hardware vs. Linux on Mainframe. I’ve read other posts here about companies buying a mainframe for stability and low maintenance cost, but until recently people didn’t buy a mainframe to run Linux. The Linux community and mainframe users made Linux work on the mainframes that people already had, to create added cost/benefit to the existing hardware. The added benefit to running on a mainframe is the true abstraction layer between the hardware and software. You can run multiple instances of linux (I’ve heard up to 40,000) on a mainframe and if any of them lock up, you kill them and start a new one.
FUD
How does Apache 2.0 stack up? It doesn’t use the same old forking model. But, it’s been said many times that Apache was never meant to be a speed demon.
On Linux, you can have static
content being served by TUX aka the Redhat Content Accelerator
and dynamic stuff such as PHP pages by Apache. I don’t know what is being done with it for Linux 2.6.
Also, can anyone offer any insight into the performance and stability of Chromium.com’s X15 webserver for Linux?
Why they don’t compare 32bit/64bit Windows on Xeons with
32/64 bit linux on Opterons ?
That is what would like too see
These benchmarks are always met with like a thousand penguinists spouting poor optimization and death threats but none of them ever seem to just shut up and deliver.
Death threats? I’m sorry, but the first overly emotional response of this thread was from a pro-MS poster, just before yours:
“God how i hate you linux zealots who don’t know jack”
This gentleman, who posts under the nickname anonymouse, then goes to make this contradictory set of statements:
“Before any of you start spewing out YOUR lies about how these benchmarks are lies, get some proof. Linux by nature is SLOW.”
I.e. he accuses others of making unsubstantiated statements, then makes one of his own…
Now, about the article at hand…has anyone actually read the report and the benchmark results? They are all about Windows 2003 vs. Win2k or NT4. Linux isn’t even mentioned once! So, really, this cannot be used as a basis to say that Win2K3 is faster than Linux 2.4/2.6, since Linux was not included in the benchmark methodology at all.
In other words, more FUD from Redmond…
Why not you compare the turtle and rabbit? The turtle lost the rush contest, i think if you were the turtle, you would complain why the rabbit has 4 agile legs?
———-Before any of you start spewing out YOUR lies about how these benchmarks are lies,get some proof.
MS has been proved to give lies and acts in the way of lies.
Just a single proof is enough to prove yours baby is wrong.
Millions of prooies still can not prove your success.
Do you want us to all act like MS, and get the same shame?
———–Linux by nature is SLOW.
Oh, dear baby, by what kinds of nature, you say this?
How about change your words to be: By nature, bill gate is my father.
The OSS community is paying close attention, especially the ones who wrote the mainframe parts of the kernel.
I know a little about mainframes, and I already know there are ways to improve Linux performance there.
BTW, one of the strong points about Linux on the mainframe is that ANY read-only file can be shared between ALL instances of Linux if it’s needed by multiple users. Another is that you don’t have thousands of servers to watch and monitor hardware for. A third is that modern mainframes have amazing failsafe features. A fourth is that mainframes are often equipped with offline backup storage that would make a PC user turn green with envy.
It would also be a real snap to work on multiple versions of a Linux system on the mainframe, for testing and other purposes, because the time it takes to make up a new VM for it is practically zero for an expert, and if the shop has tools like System Center installed, that time can be reduced to practically -1.
Yes, there are issues with Linux on the mainframe. But here, you are talking about a situation where cross-technology is not only possible, but common. For example, a truly powerful mainframe, with multiple CPUs, set up to act strictly as a database backend, running native to the hardware? Might not a little problem like somewhat-slower web page services pale in comparison to the ability of such a system to, say, process orders for goods and services? Remember that mainframe jockeys are really getting good at multi-platform environments, including virtualization of almost anything.
Microsoft is playing games with the wrong people here, I guarantee it.
News Item
Microsoft claims new benchmark for Pigs flying out of my butt and it does it cheaper than Linux.
(The preceding announcement was paid for by Waggener Edstrom, a non profit research firm based in the United States)
Occurances of the word ‘linux’ – 0
KDE and Gnome are slow…but we’re talking about servers here, not desktops. I am using a Linux file server to host about 12 gigs worth of PDFs and it is startlingly faster than the Windows 2000 server I had it on just weeks ago. Directories packed with thousands of files just “pop” open on client machines running windows. It used to take about 10 seconds. Files open faster, too. I don’t have any independent source to back up my claims, of course, but neither does MS in this case. What a surprise, MS says MS Windows is faster. What would you expect them to say? Don’t Apple benchmarks always “prove” Macs are faster than Windows? This is kind of a Duh thing…the company that put up the money for the benchmarking is going to get the benchmarks they want. Remember: Lies, damned lies, and statistics.
….when you add in the anti-virus scanning software? You know…the one that just about everyone has to have on the system these days….
Let’s just assume for a second that Microsoft is being 100% honest and Windows blows away Linux in performance. I don’t see who would care. In the small business environment, which is the only market that takes Microsoft seriously, no one cares if you can handle 2500 hits/second or 2600 hits/second. This is why people chose Microsoft to begin with. Small business will be lucky if they get 100 hits/second. The price tag, uptime, ease of license managament, and immunity to virus is really important to small business. Once the Linux expertise increases, Microsoft will find it impossible to retain customers in this segment. For the other business type, there are much better alternatives to Windows and those people know it. Unless companies enjoy rebooting their server every week or having random crashes, they will go with Sun, IBM, HP, NEC, Toshiba, Fujitsu, SGI. These are the true enterprise solution providers.
Do you think Microsoft would ever publish a report saying Linux runs better than Windows?
IIS runs partly in kernel
Yes! I can’t understand why MS did this. There are so many good reasons NOT to. In any case, I’d like to see how this stacks up aginst some of the kernel mode http servers in linux.
IBM hides in their marketing material the low performance of linux on the mainframe and MS is right, linux on the mainframe is slower than windows on a native pc, but hey the point of linux on the mainframe is NOT speed. The idea is to consolidate a ton of servers with low workload and to burn any unused mainframe capacity. Mainframes are NOT speedy computers, they are designed for monster input-output workloads and batches, not for serving dynamic html ecommerce sites and x remote KDE sessions.
If M$ says you should choose option “A” over option “B” because “A” is better for you, always choose “B”.
This is actually an endorsement of GNU/Linux on the mainframe.
Yeah … fast … ok … sure … but where’s the advantage if you have to patch it and reboot it every 10 minutes.
Micros**t’s trustworthyness is never even assumed by all in these posts. This reality stated by Linux, Apple, AND Micros**t people gives us a good picture of where we are. It is time to rid ourselves of the monopoly.
I will not use ms products in my home or business – Apple’s entry into the server market is a very good sign, as are the inroads made by Linux. If you still use ms and expect things to improve you are deluded. Why would you even complain and hand your money over to them anyway? endless viruses, security breaches….. come on. this is the 21st century! Get on with Linux or get an Apple. Otherwise you are only contributing to the problem.