Stories
Slash Boxes
Comments

News for nerds, stuff that matters

Slashdot Log In

Log In

Create Account  |  Retrieve Password

Microsoft Migrates Internal Servers to 64-bit

Posted by CowboyNeal on Fri Apr 29, 2005 07:02 AM
from the movin-on-up dept.
daria42 writes "Microsoft says servers running the company's website and MSN Search and Messenger applications have been migrated to the 64-bit version of Windows Server 2003. 'Our MSN search engine is actually built on several thousand systems running the x64 version of Windows,' a spokesperson said. In addition, 'the entire Microsoft.com site has been migrated, and we serve 30 million unique visitors every day.' According to the company, the Messenger servers handle about 70 million users."
+ -
story
This discussion has been archived. No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
 Full
 Abbreviated
 Hidden
More
Loading... please wait.
  • by nmg196 (184961) * on Friday April 29 2005, @07:03AM (#12381884)
    I almost can't believe what I'm seeing.... Maybe it's just a coincidence but I can't currently connect to MSN Messenger (Trillian crashes) AND I can't see www.microsoft.com [microsoft.com] or use Windows Update from here in the UK!

    I can't imagine that Microsoft.com could get slashdotted, so maybe they're having some severe teething issues.

    This doesn't bode well for the future of 64bit Windows computing :)
    • could just be isp/routing issues.
      Works for me, and i'm in the uk also.
    • It seems to be back now, but at the time I posted I couldn't see anything:

      Tracing route to www.microsoft.com
      over a maximum of 30 hops:

      1 1 ms 1 ms 1 ms neon.winchester.local [192.168.0.19]
      (blah)
      7 259 ms 264 ms 251 ms ten7-2.paix-osr-a.ntwk.msn.net [207.46.37.26]
      8 484 ms 263 ms 371 ms ten8-3.bay-osr-a.ntwk.msn.net [64.4.63.74]
      9 259 ms 267 ms 256 ms po2.bay-6nf-mcs-1b.ntwk.msn.net [64.4.62.138]
      10 po2.bay-6nf-mcs-1b.ntwk.msn.net [64.4.62.138] reports: Destination
      • ...for a company to "eat it's own dog food". Unfortunately in the case of MS, its software truly IS a "dog's breakfast".

        It does seem to me that the performance (or lack thereof) of MSN Messenger and related properties points to teething pains in the upgrade process. It happens quite often that you cannot sign in to Messenger or hotmail for brief periods and on some occasions you get punted. From what I have seen the problem is quite intermittent--can't sign in? Wait 15 minutes. It doesn't seem to be re
  • AMD? (Score:5, Interesting)

    by Anonymous Coward on Friday April 29 2005, @07:04AM (#12381889)
    So does this mean that it is likely that Microsoft are running AMD chips in their servers?
    • Or Intel (Score:3, Informative)

      You do realize that Intel's latest Xeons have the same AMD64 instructions too, right?
        • by Moraelin (679338) on Friday April 29 2005, @07:40AM (#12382142) Journal
          Just as a piece of trivia: Intel did want to come up with its very own 64 bit extensions, but MS basically told it that it can't be arsed to support yet another different set of 64 bit instructions. So basically the choice Intel had was squarely (A) implement AMD's set that Microsoft supports, or (B) not have any 64 bit Windows support.
    • Re:AMD? (Score:5, Informative)

      by avidday (671814) on Friday April 29 2005, @07:42AM (#12382158)
      Indeed they are - they recently bought a pile of Sun Fire V20z and V40z dual/quad Opteron servers from Sun - you can even see the Sun and Microsoft engineers posing in front of the racks here [milton.com].
  • AMD or INTEL? (Score:4, Interesting)

    by mwdmeyer (803276) on Friday April 29 2005, @07:06AM (#12381902) Homepage
    Soo, are they running on the Opteron or the new Xeon?
  • About time (Score:2, Interesting)

    by Anonymous Coward
    About time. We had a dual CPU 64-bit system back at school (between 1992-95) - some time during that time, the system was upgraded from quad 68030 to dual Risc4000 and later Risc4400 processors.

    As usual Microsoft is ten years behind times.
    • Windows was already 64 bit when the DEC Alpha came out. Which was somewhere between 1992-95 IIRC.

      The only news is is that windows now actually runs on a popular 64 bit processor. It already ran on Itanium for some time too.

  • Akamai (Score:4, Insightful)

    by A beautiful mind (821714) on Friday April 29 2005, @07:08AM (#12381917)
    'the entire Microsoft.com site has been migrated, and we serve 30 million unique visitors every day.'

    Aren't they using Akamai's help in that?
    • Yes, most of the streaming and ad content and certainly the DNS is Akamai based. Those are the 3 key services designed to take down your servers. The fact that Akamai won't run them on Windows, since it's painful to administer remotely and nearly impossible to secure, is an endless source of embarassment to Microsoft.
  • by HairyCanary (688865) on Friday April 29 2005, @07:09AM (#12381925)
    I wonder how MSN search compares to Google in terms of hardware versus load. With a couple of thousand servers in place, it would be interesting to see how many queries per second MSN search can handle per box as compared with Google...
    • by 0x461FAB0BD7D2 (812236) on Friday April 29 2005, @07:28AM (#12382062) Journal
      There's an idea for Google right there: a Google benchmark.

      Stress-test your own systems with randomized queries Google (or MSN or Yahoo!) gets and see how well it stacks up against Google's (or MSN's or Yahoo!'s) hardware, rated with GMarks (or YMarks! or....you get the idea).
    • by ScentCone (795499) on Friday April 29 2005, @07:43AM (#12382160)
      I wonder how MSN search compares to Google in terms of hardware versus load.

      Pretty much of an apples/oranges problem there, though. Yes, a search is a search is a search... but there are very different things going on relative to MSN membership, Google AdSense ads, and so on. Very different back-end processes and business issues would completely eclipse, I suspect, discussions about the individual web servers' OS. IIS on Win2003 may not be every slashdotter's cup of tea, but it's not orders of magnitude different from other servers in its ability to serve up a page. It's all that other behind-the-scenes tomfoolerly that both sites are doing that are what really weigh them down and burn up the CPU cycles. It's the database architecture and plumbing that really makes this stuff fascinating (and mysterious, if you don't work there).
      • It's the database architecture and plumbing that really makes this stuff fascinating (and mysterious, if you don't work there).

        There's an interesting video on Channel9 [msdn.com] interviewing Omar Shahine [shahine.com] that describes Hotmail internal architecture. Yup, Channel9 is a Microsoft sponsored site, and Omar is a lead program manager on a Hotmail team. He has a great blog that shows a love for devices; you'll find him talking about the iPod, Treo, PSP, etc. Channel9 also has a ton of videos on everything ranging from C#

  • And, with that... (Score:4, Insightful)

    by Stormwatch (703920) <rodrigogirao&hotmail,com> on Friday April 29 2005, @07:12AM (#12381938) Homepage
    ...they voided the computers' warranties.
  • by diegocgteleline.es (653730) on Friday April 29 2005, @07:13AM (#12381948)
    Apparently, the number of servers that run messenger went from 250 32-bit servers to 25 64-bit servers [windowsitpro.com]. Apparently it was due to a limit in the number of network connections in the 32-bit edition

    What are the "network limits" of linux, BSD, etc BTW?
    • by smitty_one_each (243267) * on Friday April 29 2005, @07:18AM (#12382000) Homepage Journal
      What are the "network limits" of linux, BSD, etc BTW?
      What do you want to program them to be today?
        • Hacking an internal product to make it work is generally frowned upon at Microsoft. If you need the app to behave a certain way, there is a good chance that other customers would too so the right thing to do is to send that feedback back to the product group so they can fix the product. Which is not to say that IT will never use special hacks to get something to work the way they want to, just that there is a resistance to doing so.
    • The limits is an performance one not code, in any of the oses you are likely to find some "reasonable default" together with a maximum setting, which could of course have been higher/use another data type if there was a use for it. I guess you can change the values for Windows settings and in some BSDs atleast you'll be limited by the maximum amount of file descriptors for the system, maybe for the user depending on settings and in NetBSD and older OpenBSDs (I think they changed it in the newer ones) a thin
  • by digitaldc (879047) on Friday April 29 2005, @07:14AM (#12381964)
    "You only need to port what's necessary," he said. "If you've got a little graphic interface and it looks real pretty and it's 32-bit, that's fine - it'll run. But when you need the 32-bit addressing, the bigger data space, certainly port that into 64-bit."
    This reminds me of some other famous quotes:

    "There are no significant bugs in our released software that any significant number of users want fixed." Source: Focus Magazine, nr.43, pages 206-212, (October 23, 1995) (http://www.cantrip.org/nobugs.html [cantrip.org])
    "Microsoft has had clear competitors in the past. It's a good thing we have museums to document that" Source: Speech at Computer History Museum (http://www.infoworld.com/article/04/10/01/HNgates talksmuseum_1.html [infoworld.com]), InfoWorld magazine, October 2001
    "640K ought to be enough for anybody."
  • by MadCow42 (243108) on Friday April 29 2005, @07:16AM (#12381978) Homepage
    But are all those systems actually running 64-bit hardware? If not, the announcement is pointless.

    MS: "Yes, our brand new car has a beautiful high-tech hydrogen fuel cell in it!"

    Driver: "But it's a diesel car..."

    MS: "Well... yes.... it's actually just sitting in the back seat for now."

    MadCow.
  • by expro (597113) on Friday April 29 2005, @07:18AM (#12382001)
    And they probably had far better OS utilization of the 64 bit architecture with their VMS or Unix. So what.

    They also had much better capability and accuracy, allowing you to search for exactly what you wanted, not just what was most popular, allowing things like the near keyword, partial word wildcarding, and many more.

    Why don't we ever hear of better search capabilities, instead of nearly-meaningless hardware shifts. The market has stagnated under Google who can't figure out how to offer even as good a search as their competitors offered at the time they launched.

    Tell me something useful.
    • Are you talking about the same Altavista the rest of us used?

      You know, altavista.com, the one that worked for a few months and then got spammed into oblivion and has been fairly unusable ever since which is why everyone now uses Google?

      I would never have described it as 'accurate'. The only reason it could possibly be seen to be accurate was because at one stage, there were no porn sites to spam the index with, so it *had* to return decent page by default - because that's all that was there.
  • by Netsensei (838071) on Friday April 29 2005, @07:21AM (#12382019) Homepage
    ... Lamborghini decided to get the engine of their next model be designed by kia
  • by MSFanBoi (695480) on Friday April 29 2005, @07:25AM (#12382042) Journal
    Before W2k was out, Microsoft migrated most internal, and everything external to W2k before it was gold. Before E2k3 was released, Microsoft was running it on all internal servers. Before W2k3 was released, Microsoft was running it on all internal and external servers. Before XP was released most workstations were upgraded to it. Microsoft has always been a very much proponent of "eating your own dog food". And yes when it goes gold Microsoft moves to that version and it's the same version sold to everyone else.
  • hotmail (Score:4, Informative)

    by OglinTatas (710589) on Friday April 29 2005, @07:31AM (#12382084)
    that would explain why my throwaway hotmail account (for recieving commercial email, and all the spam that ensues) was broken the last few days. I thought they had nerfed it again to break even more functionality in firefox and safari (they did that before) and I was just going to abandon it before I would ever load up msie. I just checked it today and it is working again.
  • Stock prices-AMD? (Score:4, Interesting)

    by va3atc (715659) * on Friday April 29 2005, @08:02AM (#12382275) Homepage Journal
    Wouldn't a move like this greatly help AMD's image?

    If its good enough for Microsoft, its good enough for us, right boss?

    Never been much into stocks, but right about now something tells me to buy. ;)
    • by gl4ss (559668) on Friday April 29 2005, @07:12AM (#12381941) Homepage Journal
      **I don't care if it's "Windows super magic XP ME 06 tournament edition". Untill you've had enough time to see how it performs for others you keep a system you know works.**

      uhh... which is EXACTLY why they're making this announcement.. so that there is "somebody" out there for it works. they're trying to boost it's acceptance you know.

    • by Psiren (6145) on Friday April 29 2005, @07:17AM (#12381995)
      This has got to be one of the dumbest comments I've ever read.

      Untill you've had enough time to see how it performs for others you keep a system you know works.

      You're saying that Microsoft, with all of its expertise at hand, is going to wait for a few other companies to roll out their OS before they do, so they can see how it goes? Give me a break. And more to the point, why would anyone else use it if even Microsoft won't. Dumb, just dumb.

    • Re:Paying with fire (Score:4, Informative)

      by michaeldot (751590) on Friday April 29 2005, @07:21AM (#12382017)

      They've got to do it. If they don't make the switch, how can they expect customers to?!

      If you read the original article [zdnet.com.au], the server is apparently quite stable (makes sense: servers run just a few processes intensively but repetitively, and cracks would show quickly), it's the client that is more questionable:

      while Microsoft is keen to tout the server version's stability, the desktop version is not as mature. Greg Sullivan, a lead product manager in the company's Windows unit told ZDNet Australia's sister site CNET News.com the desktop version "is not quite there" in terms of quality, and even hardware makers admit there might be issues.
    • by diegocgteleline.es (653730) on Friday April 29 2005, @07:22AM (#12382025)

      Anyone who is willing to switch there entire network over to something only out of beta for a few days is an idiot. It's that simple.



      In fact they've running it for months, even before the RTM date. Do you have a better way to debug the OS than putting it in servers which receive 30 millions of visits each day? (They have a farm of those to serve those 30 millions, so if one of them crashes and you lose one connection is not a big deal)

      BTW, OSDL did the same by putting linux 2.5 development versions in all their servers (getting uptimes of 200+ days in some cases BTW)
        • Hmmm...a Beowulf cluster of BSODs...

          It's not well known, but the shine of the BSODs from this cluster is the real source of the Norther Lights! ;)

    • The only way to get a BSOD on XP is to have some really broken drivers. So I'm guessing that as long as MS's servers stick to some hardware configuration known to work, they wouldn't need more reboots than any other OS does.

      I mean, let's face it, it's a server. It doesn't really need the latest ATI gaming drivers, nor a 9800 XT running at 80 Celsius just from showing the desktop, nor some experimental NForce 4 software-RAID drivers, nor a fancy sound card, etc.

      More importantly, it doesn't get all the crap
        • by Moraelin (679338) on Friday April 29 2005, @08:13AM (#12382417) Journal
          Technically it's not _needed_, and I'm certainly not going to argue with that.

          I hope you do realize, though, that it doesn't hurt either.

          1. Any library which isn't actually used, isn't even loaded. Most of Windows is just .dll files (even if some have .exe, .vxd and whatever extensions), just like most of Linux is .so files. If you don't actually run a GUI program, they won't even be loaded.

          2. Any memory page which isn't actually used, can be swapped to disc and _stays_ swapped. I.e., if after painting the desktop you don't actually run a GUI program on it, all that code to paint combos and whatnot will not even be in RAM.

          So not installing a GUI would help with... what? With the few K of RAM needed to paint the clock in the tray? (Or not even that if the taskbar is set to auto-hide.)

          And as opposed to... what? A typical Sun Solaris (UNIX) server also has all the GUI libraries, just in case you need to run some X stuff on it over the network. We have admins doing that every day. And that too means that they're loaded in memory when you do run graphics stuff, they're unloaded when you don't. Just like on Windows, eh?

          Basically what I'm saying is: before deciding that including something is dumb, please actually do an analysis, rather than just letting your ideals of perfection do the talking. You'd be surprised how much stuff may not be, technically speaking, optimal, but nevertheless is not a liability either. A lot of flame-wars could be avoided if people asked themselves "well, exactly how much does it hurt?" instead of "is it 100% perfect and 100% optimal?"
      • Any version of MacOS/X isn't actually 64 bit code, or not too much of it. It just uses some addressing extensions to be able to use more than 4 gigabytes RAM, but nothing else.

        By comparison, 64 bit Windows _is_ almost entirely 64 bit code. If you want to run 32 bit code on it, it runs in a "WOW" (Windows On Windows) virtual machine. Well, not virtual in the same way as say, Java, but in the same way as, say, Wine.
    • Re:Itanium! (Score:4, Funny)

      by roarl (137495) * on Friday April 29 2005, @07:50AM (#12382202) Homepage
      This is just a 64bit extention to a 32bit extention to a 16 bit architecture...

      No, actually. It is a 64bit extention to a 32bit extention to a 16bit extention to a 8bit extention to a 4 bit architecture. The Intel 4004 [intel4004.com] was actually the first one of this family. I guess you are too young to know.

    • Guys, just wait until Itanium is ready... This is just a 64bit extention to a 32bit extention to a 16 bit architecture...

      Opteron actually IS a 64-bit extension to the x86 hell. Same instruction set - they just extended it to 64 bit, they didn't changed anything. The success of the x86-64 architecture is being just a "extension", making very easy for compilers, software developers etc. to switch to the "new" architecture. They only added 8 registers more to the typical 8 - PPC and almost every 64-bit cp