Forgot your password?
typodupeerror
Linux Software

Sm@rtReseller and good Linux Press 93

Posted by Hemos
from the you-best-respect-me dept.
David E. Smith writes "The five millionth ZDNet publication, Sm@rtReseller, has a neat Linux article. Basically, given the same hardware, they ran three different Linux distributions and Windows NT and benchmarked them. Guess who won. "
This discussion has been archived. No new comments can be posted.

Sm@rtReseller and good Linux Press

Comments Filter:
  • Steven,

    Please don't take these anti-ZD criticisms personally. We've dealt with FUD from Jesse Berst and the likes for years. S@R, as you point out, has been very fair when doing reviews of Linux (with a few exceptions.) Unfortunately, the bulk of ZD articles comparing Linux and NT have been slightly to extremely FUD-filled. Thanks for proving that journalistic integrity still exists at ZD.

    --
    Pasc

  • It would also be interesting to see performance figures for the tested systems (as well as OS/2, and Novell) where custom tuning has been applied to the systems and applications -- runtime options in the case of IIS, compilation options and runtime options in the case of Samba and Apache.

    The value of OSS is the code. Use it!

  • by drwiii (434)
    Of course, you do know that Microsoft is paying for all this good press [zdnet.com], right?

    (Microsoft's) Maritz told Judge Jackson that a flurry of new applications and new vendors supporting Linux meant that relatively unsophisticated consumers would soon be using the operating system. As a result, he said, Microsoft would have to work diligently to keep pace with the market.

    Microsoft produced a half-dozen magazine and newspaper articles to support its contention.

  • It'd be interesting to see how much of that performance lead is due to the fact that Linux is faster than NT, and how much of it is due to the fact that Apache is faster than IIS.

    Perhaps they should try some tests with NT using the win32 version of Apache.
  • I have seen NT do the bluescreen thing with a SB16 PnP.. had to rely on the repair disk.. my kingdom for a CLI in such a situation....
  • And Linux 2.0 was a dog compared to Linux 2.2 in terms of speed. My personal Linux box is the fastest web server in all of central Florida even though the suits will never ever hire a non-Windows user here.
  • It'd be nice to do an apples to apples comparison when delivering dynamic content via CGI/PHP/ASP/Whatever, but there are too many variables to account for to give a clear picture of how well a server might respond.

    Probably the best you could hope for is a very very minimal one-liner 'hello world!' program executed in the various different platforms to check how well the web server speeds through the calling process... 'best' here wouldn't describe very accurate results of course, just the best you could hope for. ;)

    When it comes right down to it, nothing beats full scale implementation and use in the target environment... and if you REALLY want to have that comparitive execution thing going, get the same developer to create multiple versions of your program, run them all in parallel on equivalent hardware, and measure away. The numbers won't be terribly accurate, but for the price of multiply redundant programs you can have the luxury of choosing which one feels the best.

    The samba printing test would be interesting. I've run SMB in corporate environments and have had very little complaint, but it'd be nice to see how well it'd work under a heavy-duty printer workload (Say, running a linux box as a rasterizer/spooler pushing 2400dpi 11x17 lino pages in a newspaper or printhouse somewhere). Anyone in the printing industry willing to fess up to a linux rasterizer box in the closet? ;)

    --
    rickf@transpect.SPAM-B-GONE.net (remove the SPAM-B-GONE bit)

  • Can anyone explain the different Apache performance figures? I'm running Apache 1.3.3 at work for a project, and we'd like to get the best possible performance out of it. Hopefully someone will reply with "Just recompile with -02 and pentium settings" ???
    --
  • by Matts (1628)
    WebBench is freely available to download from www.zdbop.com. It contains static and dynamic content tests.
    --
  • So how does Linux *really* stack up against the Evil Empire's wanna-be-server VMS ripoff.


    "The answer: Linux with Apache beats NT 4.0 with IIS, hands down. SuSE, the least effective Linux, is 16 percent faster than IIS, and Caldera, the leader, is 50 percent faster."


    Laugh with me [zdnet.com], unless of course you're a M$ FUDster, then we'll be laughing at YOU!


    Now, how about those file server stats [zdnet.com]. ZD sums it up by saying "...Linux kicks NT's butt."


    Now, I'd like to see some other alternative OSs put against Linux. Maybe BSD?


    Anyway, it's great that this has finally come out. We all knew Linux was superior, even the M$ marketdroids. It's been a long time in the coming. Ironic, we have ZDNet to thank =].

  • Back in 1996, Infoworld compared OS/2 Warp Server with NT. They found that OS/2 was faster on one CPU than NT was on 4.

    Performance comparisons against NT always give the same result. What I would really like to see is an OEM offer two different systems for the same task, one running Linux and the other NT. Let them say that these two systems are comparable in the number of clients they can serve. And let them show that the Linux-based server is cheaper and uses older hardware. Now THAT will get people's attention.

    Unfortunately, although major OEM's are beginning to offer Linux as pre-loads for servers, I have yet to see any of them PROMOTE Linux over NT. After all, a person can buy a cheaper system if he uses Linux - and that means that OEM's can't charge a premium for top-of-the-line hardware.


    --
    Timur "too sexy for my code" Tabi, timur@tabi.org, http://www.tabi.org
  • I'd be particularly curious to see the same tests run on a tuned environment.

    Say a Proliant 6500, Quad processor, 512 Meg RAM, 100 gigs on RAID-5, etc...

    That's going to be a fairly standard corporate server, and it'd be curious to see the results with say 500 client computers.

    Unfortunately only a few places are going to have the hardware to do such benchmarking. I understand that Compaq has such a lab available.

    Also I don't believe there was mention of what kind of netbench test that was, it must have just been static content? What about dynamic content from asp or php or cgi or whatever?

    I'd also be curious how one might try to benchmark Samba as a print server. Most companies don't just install servers for file sharing, but also for printer sharing...

  • This is funny. According to law in about all civilised nations it's illegal to have an agreement that itself breaks the law. Checking your own *purchased* machine against another is something no company can forbid.

    Just imagine a car factory forbidding the owner to bechmark it against another car.

    Nobody would buy it
  • This article totally kicked my ass...I can't believe it's on ZDNet of all places! I printed out all the graphs because I'm afraid this article will disappear overnight :)

    I liked the linux-to-linux-to-linux comparisons but really wished more infor had been given as to WHY some distros perform better at certain tasks. I assume it's build differences, but I'd like to know what tricks you can use to speed up a certain program at compile-time.

    Still, this rocks! Print out the graphs and show them to your boss!

    --
  • If the GUI - X or Windows - is idle, then it doesn't suck any CPU to speak of.
  • Yes, it's always apples and oranges. Linux wasn't running a GUI, which accounts at least partly for its better performance. But, it was speaking to WinXX clients in their native language with their filesystem, while NT (as far as I know) won't do NFS. Once I heard John Blair speak and he claimed that Linux was not the best platform for Samba. All in all, I think the upshot is not that Linux is so good (pardon; it is so good) but that NT is so bad, like the comparison mentioned above with OS/2 points out. Netware doesn't do SMB, but I understand that it has the most efficient fileserver protocol, better than SMB or NFS.

    NT so often comes out last in performance compared to _anything_ else. It is difficult to understand why folks defend it, even here on /.
  • Actually, the memory used by X will be migrated to disk (swapped out) if is idle long enough, and the other processes can make use of the RAM it occupies. So really, the only thing it will end up using is some disk space...
  • If nothing else, the graphs were in the print edition of the article. (Let's hear it for free subscriptions. :) The graphs, and there's only two of them, basically show this:

    For Windows file serving, NT peaks at 8 clients and lags behind after that. Caldera peaks at 8, but much higher than NT (about 30 Mbits/sec vs. NT's 22 or so). Red Hat is basically a shadow to the Caldera line until 30 or so clients, where it is a little bit ahead. SuSE peaks at 12 or so. At the end of the graph, with 32 clients, NT has long since bottomed out at about 10 Mb/s while all three Linux distro's are still going strong in the 20-25 range.

    For Web serving, all the lines stay pretty much level across the graph. At 32 clients, Caldera is still handling 70+ requests a second, Red Hat about 60, SuSE about 50, and NT about 45.

    As others have said, Linux would have been even better if it had been tuned somewhat. (It'd be interesting to see how the NT box works when properly tuned also, just to make it a fairer fight.)

  • Sm@rtReseller is only about a year old; the current print issue, where I found the hardcopy of the same article, says "Volume 2 Issue 2." It has some interesting articles, all slanted towards biz of course. (The same issue has an article on how to configure cron, and a columnist ranting about Windows 98. Maybe there is hope for at least a small part of the soulless monolith known as ZD after all...)
  • Why would you want to go to the server room to
    do things with your server? You can have your own
    GUI on your own PC with Netscape... to administrate Samba remotely, a possibility now standard included.

    So, no need to run X on the server. Just Apache,
    which one probably would do anyway. Or an X-server on your PC. There's many ways to do a thing. But bolting a GUI to a server OS is *really* crazy.
  • That's what I did!

    And I got the contract! 100 client samba server to go, on hardware that is so much more than would ever be necessary.

  • This article makes me happy :)

    Finally, a published article truly telling the world how much Linux kicks NT's butt. Too bad they didn't tell how many times NT crashed to the gound during the tests. I bet it was more than a few.

    [drew@s196-237 drew]$ queso 127.0.0.1
    127.0.0.1:80 * Linux 2.0.35 to 2.0.9999 :)
    [drew@s196-237 drew]$
  • Follow the related links to find the graphs.

    I was amazed at the lack of FUD, the lack on install nonsense, and the generally very down to earth tone of the entire article. Did ZD just start up this Sm@rt Reseller, or just buy it? Maybe they haven't had a chance yet to destroy it. I've seen it around for a few months, which may mean ZD stuff hasn't come out of the print pipeline yet.

    --


  • So much for credibility. Go ZDNet. :/



    --
  • This article is important not because it shows
    Linux beating NT but because it compares competing
    OSS. For ZD and other commercial magazines to
    survive they must have competing products to review. If they are ever going to support OSS
    they too need a way to make money off it. Comparing different OSS products (Redhat versus Caldera, Sendmail versus Exim, or even Apache built with GCC versus Apache built with PGCC) is the way ZD can profit from OSS.
    And guess what, if they benchmark properly they'd actually be doing something useful.


  • My company recently purchased a suite of new Dell machines. NT BSOD when attempting to boot and install using the CD-ROM drive.

    Linux (RH 5.2) works fine.
  • Yeah, it was on the same PII machine (2.0.36 , glibc, vs NT4.0 SP4 with fixes and runtime from VC6.0) JDK 1.1.7a and 1.2 under NT (with JIT) vs 1.1.7a under Linux.
  • Oh please. Check out my track record--it's on the Web and in the magazine databases. I was running Unix and dissing Microsoft--when they deserve it which they usually do--for years. Check out, the Sm@rt Reseller masthead, see who's working with me in techno? A lady named Esther Schindler, better known to some of you as the OS/2 Goddess. S@R technology department calls 'em the way they see 'em. Always has, always will.

    Steven, Senior Technology Editor, S@R
  • The NT disk space number should be 1 Gigabyte and it will be Real Soon Now. And before anyone mentions that you can run it on less--or Linux either for that matter--they were all set up as network file/Web servers.

    Steven, Senior Technology Editor, S@R
  • Nah, it's rhino hide, I was just snorting and kicking at the bugs. ;-)

    Steven
  • I was just reading the NT EULA yesterday and they specifically forbid benchmarking their products without prior written approval of Microsoft. I wonder if it was obtained for these tests...
  • It also depends on what cpu they compiled it for. Redhat compiles with no cpu optimization, i.e. no -i486, -mpentium, etc. I think that that was what a new dist called somethin glike Stampede was partially about, compiling everything without debugging symbols and for pentiums. Those CPU optimizations can significantly affect performance. That being said, I don't know what cpu optimizations each dist uses. Also, it's not a given that everyone optimizes with -O, -O1, -O2, -O3, etc.
  • I don't know what NT does on SMP, but Linux 2.0.x uses BFL, big grained SMP, which is a big disadvantage. 2.2, by contrast, should really look nice on SMP systems, especially bigger ones. Of course, 2.2 dists aren't going to be out for a few months, so that will take a while. It will be fun to see, though, especially given Apache's multi-process architecture.
  • On the one hand you can be sure that MS is looking for articles that support their contention that they are not a monopoly.

    On the other hand they don't want to add any legitimacy to any real competition.

    I don't think this article would do their case any good. How can you explain the ability to charge $800+ for a file server that is outperformed (2.5x) by a $50 linux distribution!

    MS is in a position where they want to portray Linux as immature potential future competitor. To do this they don't need any articles that actually include facts. They want articles that talk about how great it is if you are a Unix genius but otherwise it's impossible to install.

    CM
  • There was a table showing which version of Apache for each distribution. They apparently used the distribution's straight without any recompiles or anything. A couple of the versions were 1.30 and one was 1.31. (See the link on how they stack up).
  • Or more likely just VM (if the other tasks are more active).
  • Quoting article:
    "Where Linux does fall behind, however, is driver support and hardware discovery. Although Linux hardware support is improving, you must check each version's supported hardware list to be certain that all of your customers' components will function properly."

    The same is true of NT, I saw a video card installed in an NT system thet caused a bluescreen every time you booted. Not only that, it hosed the install and required a complete reinstall, every time. The installer later found out thet NT did not support the card. Unfortunately I do not remember what video card it was.


    "Trouble is, just because it's obvious doesn't mean it's true"

  • It's a feature of the OS that allows you to tune it to what you are doing. Besides, even if you did run X as long as you had nothing that required X updates running I doubt it would really make that much of a difference (though I could be completely wrong here).

    "Trouble is, just because it's obvious doesn't mean it's true"

  • Sm@rt Reseller has been fairly decent for a long time. They had a biased pro-Microsoft opinion article a while back about the Microsoft vs DOJ trial, and when I pointed it out, and wrote a rebuttal it took only a few hours before I got a really excited mail from the editor that wanted to print it... :) And I wasn't very nice to the guy who wrote the original article.

    In the few cases where I've seen FUD filled articles there, it has been opinion pieces, clearly marked as such, and they have been quick to include articles representing the opposing view.

    Imagine, actual journalism from a ZD publication :-)

I wish you humans would leave me alone.

Working...