VeriTest, an independent test lab, has found that Windows Server 2003 outperforms Microsoft Windows 2000 Server and Microsoft Windows NT Server 4.0 by a dramatic margin—typically performing two to three times faster on the same hardware. Compared to Windows NT Server 4.0, Windows Server 2003 is: Two times faster on average as a file server. Three times faster serving dynamic Web content. Four times faster serving static Web content. Update: Apparently, VeriTest is owned by Lionbridge.
Lab Report: Windows Server 2003 Outperforms Predecessors
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
I like the fact that the windows server had its drives mounted with noatime, whereas the linux hadn’t. Who knows what else might be “tweaked” a little on the windows side?
But nontheless impressive performance.
I would rather see a completely independent test, not one ‘commissioned by Microsoft’. It’s not that I doubt the quality of the OS, I just don’t trust the results when MS commissions the study and writes a big check to get the results.
If you go to the store to buy a computer and ask “I’ve already have an OS, could I buy it without OS?” they’re likely to reply “Nope, Sorry”. What a shame! How many computers have you seen without the “designed for Windows” sticker? My IBM didn’t come with one
and never used the Windows-based recovery CD. I just don’t trust things based on a system which spies you, don’t know what’s running on the background and is the opposite as Slackware’s great SSS -simple, stable, secure-. XDD
n0dez
I dunno but how independent is this test? For example check this link: http://www.veritest.com/clients/reports/microsoft/ms_netbench.pdf
I refuse to believe that win2003 is faster then samba on RH. Oke maybe by 20%… maybe. but +66 till 95% thats impossible. In tests also see bad things, well I don’t see this.
IMHO this is pure PR
“Three times faster serving dynamic Web content” Yowza, I wonder if I can ship a few of our Xeons back and use the money to upgrade Hydra, our robot.
Seriously though, it’s amazing how inefficient 2K Server must have been if it’s getting beat down like it is by it’s younger sibling. Maybe, just maybe, MS had a deal with the hardware vendors that would keep it’s software inefficient to boost hardware sales. Didn’t much matter to MS back when there was no competition.
From what I have read about Windows Server 2003 it that it is being marketed as an upgrade to Windows NT 4 server. And if that is the case I’m guessing that performance on older hardware is more important.
you still need 50 servers for 100 workers (yeah yeah it is an exageration but you get my point)
So the next CodeRed , nimda YAworm will spred 5 times faster!
Is the fastest OS ever. It will blow away both OSX and Linux. Just like XP did.
win server 2003 is faster than old servers?
That is generally the idea, the new product should be better than the old.
New MS licenses are expensive and restrictive. You may require new hardware. Rather than switch to these go with Linux. It is good enough for Wall Street’s Masters of the
Universe it is good enough for you. While we can debate linux on the desktop there is no doubt that Linux kicks Windows butt in the server space.
I haven’t used Windows since long time ago. However, both Slackware and FreeBSD run pretty fast.
n0dez
To make linux look as bad as possible in a benchmark, use ext3. It has data journaling and big kernel lock.
If performance is required, xfs or reiserfs (depending on what you want) would be betterd suited.
I couldn’t find any mount options in the pdf so data=ordered (meta data journaling) is being used, but who knows.
Perhaps this reference tells something about the status of ext3:
http://www.ussg.iu.edu/hypermail/linux/kernel/0206.2/0853.html
/jarek
Dave Cutler did an excellent job with VMS.
But with WNT he did an even better job.
Now all Nixes are visually degraded to what they always have been, toy operating systems of idiots.
Guess what?
UNIX/Linux is histoty.
Just look at the latest TPC benchmarks.
WNT rules the 64 bit world.
how does it compare to netware 4.x or better?
The world seems to be switching en masse to Sun and other forms of UNIX-like setups.This lab data would be of greater use if it compared the NT scores (or server 2003, whatever) to traditional types like System V, NetWare or FreeBSD which Microsoft itself uses.
Anyway, great job, MS!! I hope somebody buys it. Heck, I think that a fully licensed UNIX setup is cheaper than MS now
The performance of Windows Server 2003 compared to RedHat sounds like the results of a presidential election in a corrupt country : candidate A got 86 % of the votes in province X, 95 % in province Y and 100 % in province Z.
Everybody, stand up and cheer for our new elected leader, the Great One, Spyware Server 2003 !
It looks as if the fix was in, along with a thick briefcase full of M$ cash
i tried win2003 on an old server it was a dual p3 450 with 256mb ram
and a ibm 10krpm uw disk. but i must say that i dint beanch it it was to slow on every thing on the desktop it felt relly slow. ok it a server win2k was really snappy on that machine. i was so tierd of siting in front of it and configuer things when everything lagged so i installed gentoo insted. hmms i have been thinking about doing som bench test betwen gentoo and w2003 server on that machine later. will se if i get the time
That’s it everyone, keep ranting =)
When you add the required security fixes it will slow down considerably. Let the patching begin
Whats wrong? At least you can still claim to be smarter than the BeOS for Enterprise? people. (Not by much though since even they have better ACLs than Linux.)
What does this have to do with servers? Most servers, especially web servers, don’t already come with Windows in the first place. Besides, the other day I went into an Apple store and asked “I already have an OS, can I buy a Mac without the OS?” can you guess the response?
Well, i mean this was fair to compare the new Microsoft 2003, with some really optimized Linux distro. Because Gentoo _is_ slower, and by a wide range, than RH…
ok it a server win2k was really snappy on that machine. i was so tierd of siting in front of it and configuer things
Why not ? I can’t draw what it is exactly your point when you speak about “too much stteing tasks”, but let say that is a matter of taste…
when everything lagged so i installed gentoo insted
Needless to say, your whole stuff become here incomprehensible.
Something like Lindows or Xandros or another Lycoris, may have been meaningfull, but Gentoo, a distro which required, at least, one _week_ ( i mean a trully 24/7 running session… ) just to have a poor empty desktop manager installed, if you are lucky…
Is that your solution to simplify your setting task ?
Not to speak of the “little” Gentoo problems. Just the funiest of them :
* xfree and TK are _required_ to emerge a simple kernel ( if don’t know this distro, Gentoo force you to compile a kernel, without that : nothing ! And, of course, the kernel is not parameterized out of the box… ). Ho yeah linux trolls, portage avoid the libraries hell. Well, try emerge –nodeps, it’s more simple ;-)))
* anyway, xfree doesn’t compile ( some strange library are missing in the ebuild ). Ho yeah, portage avoid libraries hell ;-)))))
* the same for TK. The very wise 8.3 release, not really the state of the art ;-)))
* to finish, a little question for our simplicity king : how do you use the “wonderfull” portage to compile a package with some specific option ? Ho no… Do you really mean that you have to break out of portage and get back on the very old tar.gz stuff, and very manually setting an emerge –inject for each part of the package, to correct the wonderfull portage tree for an installation it _can’t_ manage
Eventually you’re right, Gentoo is obviously the simpliest solution… in linux trolls land ;-))))))))))
This platform is obsolete. The only way to go is Linux because of the low cost and high security.
doesn’t paying a company 3 million dollars kinda defeat the purpose of objective evaluation of a product? in any other arena, it would be called bribery… when it ocmes to MS, it called status quo
> * xfree and TK are _required_ to emerge a simple kernel…
From /usr/src/linux/README:
– Do a “make config” to configure the basic kernel. “make config” needs
bash to work: it will search for bash in $BASH, /bin/bash and /bin/sh
(in that order), so one of those must be correct for it to work…
– Alternate configuration commands are:
“make menuconfig” Text based color menus, radiolists & dialogs.
“make xconfig” X windows based configuration tool.
“make oldconfig” Default all questions based on the contents of
your existing ./.config file.
Yes, if you want to use ‘make xconfig’, you’ll need X, but if you’d bothered to read the documentation you’d have realized this and understood that it’s only one of three options and that complaining about it only makes you look like an arse.
.. and so few software benchmarks? Anyway, it does look like MS is making major progress. A year ago Linux advocates were saying Linux would be taking over the world about now.
For Mission critical, vital business computing.
I waiting for the fair benchmarks… you know the real tests like:
“Gentoo vs. Win95 as an application server”
“Apache 1.7 vs. IIS 1.7”
“Mandrake 9.0 vs. WinXP Home for K-Apps support”
“SuSE 8.0 vs. WinME for number of german software developers involved”
“Slackware vs MS-DOS 5.0 in launching OO.o performance”
“RedHat vs Windows 64bit Server for UltraSparc support”
Even if NTFS (or whatever filesystem is now being used) was slow and horrible like its previous bretheren, I doubt it would be so slow to cause excessive IO blocks for file serving vs. those in the network stack. Even Winblows caches filesystem IO in memory. Given similar hardware, and assuming network drivers and the TCP stack can’t be optimized THAT much, how could fileserving POSSIBLY be THAT MUCH improved? Were there sleep() calls in earlier IOs? This makes no sense. I can absolutely saturate a network with some of the slowest hardware and the slowest OSes. Is SMB that convoluted?
—
3lixyqueue
1) Why are there flames over this? this is just another PR stunt. Companies who have a clue about purchasing computer equipment do not simply look at the latest “PR spin” from MIcrosoft and take it as evidence. Most companies I have been with evaluate it with in the own organisation to see whether or not the investment, in their circumstances gives them the increase in performance they are looking for. Once they find that information out, they compare it to the costs and decide whether to stay, upgrade or migrate.
2) I’ve read the reports and IMHO, they either say two things, 1) The study is a load of crap or 2) HP, IBM, Shell, BP, Boeing, Damler Chrysler and numerous other companies have been taken for a ride on the Linux hype. Personally I find that study is a load of crap ordered by a company despirately grasping at straws as they realise their network of suckers in high places willing to use Microsoft no matter what, are now geting the strong word by the shareholders to cut expenses. It is about time Microsoft started looking at WHY people are moving away from their products. Simply taking the sun defence and blaming everyone else isn’t going to help.
interesting im glad osnews didnt buy into the bullshit.
“Update: Apparently, VeriTest is owned by Lionbridge.”
HA HA HA that says it all..
so microsoft try harder next time..
im frankly getting tired of the to and fro bashing bash away as much as you like on desktops where linux is becoming a viable threat. but dont get into the server market.. osnews please stop posting every little crappy development about this os its just getting boring.. they dont stand chance ill explain why.. A SERVER DOES NOT NEED A GUI unless its forwarding a gui to dumb clients in which case ms fails miserably anyway. its that simple. also a server does not need bloat should not be rebooted every time new dlls are installed everytime its patched every time a new piece of software that adds more dll is introduced.
Please stop this shit with win 2003. the arguement is dead ms can produce fud from here till 10 years time who believes their bullshit and uses *nix literally none of us. for it to be proven to be good needs a lot more it just wont get the nix heads to convert so it can only increase its tight hold on the market space it already has and boast about the new adoptions from nt4 server and 2000 server. thats it end of story..
can you see a company giving up its sun solaris clustering solution to replace it with windows ? dont make us laugh..
as for dependancy hell etc etc..
come on people get with the program how long has debian been around for ? yeah gentoo is a new comer and sorts out dependency hell but how long has debian been around for ? and apt-get sorted that problem years ago well done to redhat for finally incorporating apt-get into rpms and congrats to who ever actually did it.
i love the way right at the top it says:
“VeriTest, an independent test lab” bull shit.
sorry but more fud again from the great fud maistros at microsoft marketing inc.
microsoft should go into politics..
one good thing microsoft do make though is bags. maybe they should give up the software market and make ruck sacks. seriously my mates got one he got as a freebie my god the bag is awesome.. first time youll hear me call an ms product awesome lmao
I think this is a hint and pitch for the still large mass of NT4 users which don’t want to switch. If it is going to be heavily promoted as a drop in NT4 replacement (i.e. “runs faster on the same machine”) than probably a part of these users will switch so M$ can have them on the new crop of OSes. Probably for those NT4 users not upgrading the big server running NT4 but only the OS would pe a good value proposition from M$.
Probably by now many looked at Linux and at least heard of it. I think M$ has a tougher time selling his server software than before.
I cannot understand Microsoft. Windows 2003 doesn’t run on NT4 hardware. Why in the heck is their marketing department pushing Windows 2003 on the NT4 crowd?
NT4 is well known, debugged, and has the least spyware in it of all the current Microsoft operating systems. If it weren’t for Microsoft updating their software not to run on NT4, I’d run all my Windows servers on NT4.
A corporation that has a working NT4 system would be stupid to switch to any of Microsoft’s latest systems, all of which are buggier, take more hardware to run, cost much more to run and contain an order of magnitude more spyware.
Even if NTFS (or whatever filesystem is now being used) was slow and horrible like its previous bretheren,
You mean OS/2 HPFS then?? One can say alot about MS and it’s software/business methods, but NTFS is actually a quite nice piece of work
$899 worth. I have a copy of NT4 that still is almost as good and I didn’t pay that much for it. I would go for a WindowsXP that supported 4 processors. What a goofy deal. They are selling dual xeon’s but you have to pay a grand to MS to run your new machine. Linux 2.sumthing runs all four.
What spyware?
> but NTFS is actually a quite nice piece of work
Yes, it is, but that’s what I was saying, “even if NTFS…was slow…” and it’s not, I’ve seen NTFS perform VERY well. Precisely my confusion.