Quoting: Is Windows 2003 Server really faster than Linux/Samba? Here comes Windows 2003 Server! And it's faster than Linux, because Microsoft cheated with the benchmarks! May 9, 2003 Article summary: Microsoft claims Windows 2003 Server is twice as fast as Linux, at least when it's used for file serving. I spoke to Jeremy Allison, head of the Samba team, who provided a few insights into the test configurations that don't leap out at the reader because they are hidden away in appendixes to the benchmark document. Allison feels this, in itself, is substantially responsible for the outcome. (1,000 words) See: http://www.linuxworld.com/go.cgi?id=742352 -- Fred A. Miller Systems Administrator Cornell Univ. Press Services fm@cupserv.org, www.cupserv.org
On Tuesday 13 May 2003 16:36, Fred A. Miller wrote:
Quoting:
Is Windows 2003 Server really faster than Linux/Samba?
Here comes Windows 2003 Server! And it's faster than Linux, because Microsoft cheated with the benchmarks! May 9, 2003
Article summary:
Microsoft claims Windows 2003 Server is twice as fast as Linux, at least when it's used for file serving. I spoke to Jeremy Allison, head of the Samba team, who provided a few insights into the test configurations that don't leap out at the reader because they are hidden away in appendixes to the benchmark document. Allison feels this, in itself, is substantially responsible for the outcome. (1,000 words)
What I interpreted from that is something that I have been saying for a long time and that is Windows is better suited for SOHOs, mom-n-pop shops, than Linux. From the user perspective it is pretty obvious in that most people are used to the Window GUI and the switch to KDE or Gnome is an extra burden that costs any company and resources. Now we have the server perspective. It seems that right out of the box Windows is faster than Linux. Isn't that what the mon-n-pop shops want? They want something they can quickly install and not spend the time tweaking. AFAIK the ext3fs is default for many distributions and not XFS. So the out of the box (i.e. the default) Linux has some weaknesses. In a company with it's own IT department, where they know about things like this (or should know), the can make the neessary changes to the default (even without "tweaking"). Regards, jimmo -- --------------------------------------- "Be more concerned with your character than with your reputation. Your character is what you really are while your reputation is merely what others think you are." -- John Wooden --------------------------------------- Be sure to visit the Linux Tutorial: http://www.linux-tutorial.info --------------------------------------- NOTE: All messages sent to me in response to my posts to newsgroups, mailing lists or forums are subject to reposting.
James Mohr wrote:
On Tuesday 13 May 2003 16:36, Fred A. Miller wrote:
Quoting:
Is Windows 2003 Server really faster than Linux/Samba?
Here comes Windows 2003 Server! And it's faster than Linux, because Microsoft cheated with the benchmarks! May 9, 2003
Article summary:
Microsoft claims Windows 2003 Server is twice as fast as Linux, at least when it's used for file serving. I spoke to Jeremy Allison, head of the Samba team, who provided a few insights into the test configurations that don't leap out at the reader because they are hidden away in appendixes to the benchmark document. Allison feels this, in itself, is substantially responsible for the outcome. (1,000 words)
What I interpreted from that is something that I have been saying for a long time and that is Windows is better suited for SOHOs, mom-n-pop shops, than Linux. From the user perspective it is pretty obvious in that most people are used to the Window GUI and the switch to KDE or Gnome is an extra burden that costs any company and resources.
Now we have the server perspective. It seems that right out of the box Windows is faster than Linux. Isn't that what the mon-n-pop shops want? They want something they can quickly install and not spend the time tweaking. AFAIK the ext3fs is default for many distributions and not XFS. So the out of the box (i.e. the default) Linux has some weaknesses. In a company with it's own IT department, where they know about things like this (or should know), the can make the neessary changes to the default (even without "tweaking").
Regards,
jimmo
If you read the report, the Windows server wasn't an out of the box install. It had been tweaked using various registry hacks, much out of the league of mom-n-pop shops. They also used a 64K file system block size, whereas, IIRC, the default for NTFS is 4K, just as with the ext3. The whole point is that what was tested wasn't representative of what would be the default for a Windows server installation. So the comparison isn't valid. Steve
On Sunday 18 May 2003 16:10, Stephen Allewell wrote:
If you read the report, the Windows server wasn't an out of the box install. It had been tweaked using various registry hacks, much out of the league of mom-n-pop shops. They also used a 64K file system block size, whereas, IIRC, the default for NTFS is 4K, just as with the ext3.
The whole point is that what was tested wasn't representative of what would be the default for a Windows server installation. So the comparison isn't valid.
Steve
A mom-n-pop shop doesn't care about kernel/registry tweaks. All they care about is ease of use and the "marketing" aspect of Windows being faster (valid or not). Microsoft plays down the fact that there are tweaks involved, just that it is "faster". If you only look at the first few pages of the report (which most people do anyway) you don't see anything about this. So for mom-n-pop, Windows is faster. It's not until you read the details (which mom-n-pop don't care about) that you find out the details. Sorry, I seriously doubt that mom-m-pop know what a block size even is, let alone what effect it has on the performance. The only effect it does have is to make them think Windows is faster. Windows "performance" is all in the marketing and thus better suited for the mom-n-pop shops. Really quite logical when you look at it. Regards, jimmo -- --------------------------------------- "Be more concerned with your character than with your reputation. Your character is what you really are while your reputation is merely what others think you are." -- John Wooden --------------------------------------- Be sure to visit the Linux Tutorial: http://www.linux-tutorial.info --------------------------------------- NOTE: All messages sent to me in response to my posts to newsgroups, mailing lists or forums are subject to reposting.
James Mohr wrote:
On Sunday 18 May 2003 16:10, Stephen Allewell wrote:
If you read the report, the Windows server wasn't an out of the box install. It had been tweaked using various registry hacks, much out of the league of mom-n-pop shops. They also used a 64K file system block size, whereas, IIRC, the default for NTFS is 4K, just as with the ext3.
The whole point is that what was tested wasn't representative of what would be the default for a Windows server installation. So the comparison isn't valid.
Steve
A mom-n-pop shop doesn't care about kernel/registry tweaks. All they care about is ease of use and the "marketing" aspect of Windows being faster (valid or not). Microsoft plays down the fact that there are tweaks involved, just that it is "faster". If you only look at the first few pages of the report (which most people do anyway) you don't see anything about this. So for mom-n-pop, Windows is faster. It's not until you read the details (which mom-n-pop don't care about) that you find out the details. Sorry, I seriously doubt that mom-m-pop know what a block size even is, let alone what effect it has on the performance. The only effect it does have is to make them think Windows is faster. Windows "performance" is all in the marketing and thus better suited for the mom-n-pop shops.
Really quite logical when you look at it.
well....this all depends on how you look at it. Microsoft used a similar approach against Novell a few years back making winNT look faster on a network compared to a novell server. They purposely made windows listen for requests for novell servers, when they picked that up, they would put the the novell requests into a loop - not a continous loop, but a long enough time that the typical end-user would sense the delay. To the un-informed person.....windows seemed faster! This was documented in the anti-trust trial. Novell made MS aware of this immediately...now long did MS take to fix this? Approx. 2 years!! (not a coincidence!) By then, the damage was done.....people are so gullible! This is the kind of bullshit I hate about microsoft...these tactics they use! We must do a better job of educating people and making them aware of the truth! To me, the performance comparison is a fraud. It manipulates things to look a certain way, the way MS wants it to appear. But it's not the truth! I also don't like the fact the current american gov't (Bush admin) has turned a blind eye to MS's bullshit......I guess it's easy to do that when MS "contributes" [bribes, in my opinion] millions of dollars to the republicans! Last stat. I saw was around $4 million ..... so that's what it costs to "pay off" the anti-trust case...hmmmm.
Regards,
jimmo
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Sunday 18 May 2003 13:37, Oskar Teran wrote: <snip>
well....this all depends on how you look at it. Microsoft used a similar approach against Novell a few years back making winNT look faster on a network compared to a novell server. They purposely made windows listen for requests for novell servers, when they picked that up, they would put the the novell requests into a loop - not a continous loop, but a long enough time that the typical end-user would sense the delay. To the un-informed person.....windows seemed faster! This was documented in the anti-trust trial. Novell made MS aware of this immediately...now long did MS take to fix this? Approx. 2 years!! (not a coincidence!) By then, the damage was done.....people are so gullible!
I agree 100%.
This is the kind of bullshit I hate about microsoft...these tactics they use! We must do a better job of educating people and making them aware of the truth! To me, the performance comparison is a fraud. It manipulates things to look a certain way, the way MS wants it to appear. But it's not the truth! I also don't like the fact the current american gov't (Bush admin) has turned a blind eye to MS's bullshit......I guess it's easy to do that when MS "contributes" [bribes, in my opinion] millions of dollars to the republicans! Last stat. I saw was around $4 million ..... so that's what it costs to "pay off" the anti-trust case...hmmmm.
It isn't *this* administration, it's been *ALL* the administrations that have turned blind eye*s* on M$' greedy, monopolistic, and unscrupulous tactics and business practices. M$ has been "bribing" almost since day one, so don't go blaming any one administration. Clinton had *two* terms and left one of the biggest messes this country has ever had to clean up. There, I got to put in my little political beliefs. pbtpbtpbtpbtpbt!!!
Regards,
jimmo
John - -- A butterfly is: Pretty,soft,harmless...and useless, just like M$N. My Penguin and my Gecko eat butterflies. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.2-rc1-SuSE (GNU/Linux) iD8DBQE+yFPZH5oDXyLKXKQRAuJOAJ96aatn1pevi4XUqeT0ZDogsIZM9wCcDL1+ QQGpHcozvD0QnPQohU5vRIs= =BFuT -----END PGP SIGNATURE-----
James Mohr wrote:
A mom-n-pop shop doesn't care about kernel/registry tweaks. All they care about is ease of use and the "marketing" aspect of Windows being faster (valid or not). Microsoft plays down the fact that there are tweaks involved, just that it is "faster". If you only look at the first few pages of the report (which most people do anyway) you don't see anything about this. So for mom-n-pop, Windows is faster. It's not until you read the details (which mom-n-pop don't care about) that you find out the details. Sorry, I seriously doubt that mom-m-pop know what a block size even is, let alone what effect it has on the performance. The only effect it does have is to make them think Windows is faster. Windows "performance" is all in the marketing and thus better suited for the mom-n-pop shops.
Really quite logical when you look at it.
Regards,
jimmo
It just goes to prove that perception is everything, even when there is no substance. Statistics can be manipulated, test systems can be engineered to produce the desired result. If mom-n-pop are taken in by this, reading a couple of pages of a dubious report, without further investigation, then they are safe in their own ignorance, that they have a fast system that suits their need. I wouldn't have thought that Windows server 2003 was really aimed at this market anyway. So there is hope that the more technically inclined people who are looking at this report would at least be able to sift out the dross and look at the facts instead of just the hype. I would be interested to find out why the hacks that were used are not part of the default installation, perhaps they are a bit risky for a production setup, or more likely to hurt performance in a real world environment! Steve
I wouldn't have thought that Windows server 2003 was really aimed at this market anyway. So there is hope that the more technically inclined people who are looking at this report would at least be able to sift out the dross and look at the facts instead of just the hype.
Sadly, more techicanlly inclined people are as easy as everyone else to cheat in my experience. Some guys just do not have any criticism on what is under their eyes, even if it hurts them. To be short: many techincally inclined guys are idiots, just like the others.
I would be interested to find out why the hacks that were used are not part of the default installation, perhaps they are a bit risky for a production setup, or more likely to hurt performance in a real world environment!
The 64kbyte block size makes me think that a configuration like that would not work very well in a real world environment, as it would waste a lot of space. But it is just good when you are trying to fill the pipe like in that test. I am sure they do not have tryed to move a thousand of small files, right? Praise
Praise wrote: <snip>
The 64kbyte block size makes me think that a configuration like that would not work very well in a real world environment, as it would waste a lot of space. But it is just good when you are trying to fill the pipe like in that test. I am sure they do not have tryed to move a thousand of small files, right?
Praise
The report doesn't say how big the files being served were, but certainly the bigger they were, the more likely the bigger block size would have helped the overall throughput, and that was really all that was being measured. I have just reread the report and loved the disclaimer at the end. An extract below. VERITEST SPECIFICALLY DISCLAIMS ANY WARRANTY, EXPRESSED OR IMPLIED, RELATING TO THE TEST RESULTS AND ANALYSIS, THEIR ACCURACY, COMPLETENESS OR QUALITY, Steve
On Sunday 18 May 2003 21:28, Stephen wrote:
James Mohr wrote:
A mom-n-pop shop doesn't care about kernel/registry tweaks. All they care about is ease of use and the "marketing" aspect of Windows being faster (valid or not). Microsoft plays down the fact that there are tweaks involved, just that it is "faster". If you only look at the first few pages of the report (which most people do anyway) you don't see anything about this. So for mom-n-pop, Windows is faster. It's not until you read the details (which mom-n-pop don't care about) that you find out the details. Sorry, I seriously doubt that mom-m-pop know what a block size even is, let alone what effect it has on the performance. The only effect it does have is to make them think Windows is faster. Windows "performance" is all in the marketing and thus better suited for the mom-n-pop shops.
Really quite logical when you look at it.
Regards,
jimmo
It just goes to prove that perception is everything, even when there is no substance. Statistics can be manipulated, test systems can be engineered to produce the desired result. If mom-n-pop are taken in by this, reading a couple of pages of a dubious report, without further investigation, then they are safe in their own ignorance, that they have a fast system that suits their need.
That's what's marketing is all about anyway. If Britney likes Pepsi, then it must be the best soft drink, right? If a company that calls itself "VeriTest" says that they tested the two systems and NT was better, it has to be right.
I wouldn't have thought that Windows server 2003 was really aimed at this market anyway. So there is hope that the more technically inclined people who are looking at this report would at least be able to sift out the dross and look at the facts instead of just the hype.
I guess it depends on your definition of "mon-n-pop", but I generalize as use that term for small companies that really don't have their own IT department. Unfortunately the management making the decisions are often the ones reading the same first few pages. They are also the ones making the decisions about "re-training". Comments I have received about the "re-training" only taking a few hours are unrealistic. First companies need a real, tangible reason to switch and spend all that money on re-training. Even if are talking about "only" 20 people. "Proof" that Veritest cheated is not tangible. On the other hand, current MS license policies **is** something tangible that would encourage then to switch.
I would be interested to find out why the hacks that were used are not part of the default installation, perhaps they are a bit risky for a production setup, or more likely to hurt performance in a real world environment!
I would imagine that it has to do with some concept of what the "average" installation is. The configuration they used is probably not as efficient in every case. The defaults are sufficient for most functions, but you need to tune it for your specific applications. For example, an online transaction processing system has different requirements than a filesystem with large text documents. Regards, jimmo -- --------------------------------------- "Be more concerned with your character than with your reputation. Your character is what you really are while your reputation is merely what others think you are." -- John Wooden --------------------------------------- Be sure to visit the Linux Tutorial: http://www.linux-tutorial.info --------------------------------------- NOTE: All messages sent to me in response to my posts to newsgroups, mailing lists or forums are subject to reposting.
participants (7)
-
Fred A. Miller -
James Mohr -
John -
Oskar Teran -
Praise -
Stephen -
Stephen Allewell