OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
From: Charles Gamble (Charles.GambleSINGULARITY.CO.UK)
Date: Mon Oct 08 2001 - 10:00:48 CDT

  • Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

    One thing we have just noticed is that the Dllhost.exe which is hosting our
    com+ dll's is page faulting extrememly excessively
    ( average of 700 page faults per sec when procesor object charted in
    perfmon) in Com+ .
    In Nt 4.0/Mts it is no where near this figure.
    Any ideas what could be causing this?
    Our threading model is "Both" for all our objects.

    Charles Gamble.

    -----Original Message-----
    From: Richard Blewett [mailto:richard.blewettVIRGIN.NET]
    Sent: 03 October 2001 16:06
    To: DCOMDISCUSS.MICROSOFT.COM
    Subject: Re: COM+ and Performance

    Sorry, I didn't mean the COM+ subsystem, I mean't that the OS in general -
    disk I/O etc was faster in Win2K than NT. There are loads of factors which
    can affect your performance when running under COM+ and straight porting
    from MTS to COM+ isn't necessarily an optimal way to go. The context
    architecture and threading issues are much more complex under COM+ (theres
    alot more choice for starters).

    Richard
    http://staff.develop.com/richardb

    -----Original Message-----
    From: Distributed COM-Based Code [mailto:DCOMDISCUSS.MICROSOFT.COM]On
    Behalf Of Christine Wolak
    Sent: 03 October 2001 15:36
    To: DCOMDISCUSS.MICROSOFT.COM
    Subject: COM+ and Performance

    We too have been having performance problems with our application on
    Windows 2000. Here's a summary of where we are now, in case it helps
    anyone:

    Our application was taking about 2.3 times as long under Win2k as compared
    to NT. We got a hotfix from Microsoft which got us about 0.6 of that 2.5
    back. This hotfix was supposed to provide MTS emulation mode (as far as
    max apts per thread, or max threads, or something like that), but the
    emulation mode itself didn't make any difference in our application - just
    having the hotfix did.

    Then, all of a sudden we got all but 0.1 of it back. A significant
    improvement! The traces we did pointed to the function
    GetQueuedCompletionStatus, which seemed to have to do with I/O in some
    way. I think I think this improvement came because of
    manually/specifically stopping and restarting DTC (rebooting the machine
    did NOT do the trick). Microsoft never did say for sure why or whether
    stopping DTC fixed this, what might have caused it to take up so much time
    all of a sudden, how to keep it from happening, etc. We haven't seen the
    problem crop up again, but if we do, Microsoft says that they'll want to do
    some poking around at call stacks and stuff.

    Another thing that did not apply to us but may apply to others was a memory
    allocation issue that was broken in SP1 but fixed in SP2.

    We also did some testing to force everything into ONE BIG CONTEXT, and that
    had no real impact on our performance. We have one big
    package/application, though, so maybe that has something to do with it.

    I must respectfully disagree with Richard Blewett, who said "for a properly
    configured server, Win2k is simply faster than NT". This is NOT the case
    for every COM application, and as far as I could tell from Microsoft,
    there's not much in the way of "configuration" of your server that you can
    do to speed up performance, anyway. Maybe for WELL-DESIGNED systems, or
    small systems (we have 71 DLLs in COM+), or some other scenario.
    But "configuration of the server" didn't seem to have anything to do with
    it.

    ----------------------------------------------------------------
    Users Guide http://discuss.microsoft.com/archives/mailfaq.asp
    contains important info. Save time, search the archives at
    http://discuss.microsoft.com/archives/index.html .
    To unsubscribe, mailto:DCOM-signoff-requestDISCUSS.MICROSOFT.COM

    ----------------------------------------------------------------
    Users Guide http://discuss.microsoft.com/archives/mailfaq.asp
    contains important info. Save time, search the archives at
    http://discuss.microsoft.com/archives/index.html .
    To unsubscribe, mailto:DCOM-signoff-requestDISCUSS.MICROSOFT.COM

    ----------------------------------------------------------------
    Users Guide http://discuss.microsoft.com/archives/mailfaq.asp
    contains important info. Save time, search the archives at
    http://discuss.microsoft.com/archives/index.html .
    To unsubscribe, mailto:DCOM-signoff-requestDISCUSS.MICROSOFT.COM