RE: Windows 2003 Server throughput vs. Windows XP

Hello Jeffrey,

I have been doing a lot of other investigations around this problem, and the
same problem shows on desktop Pentium 4 computers with hyperthreading
disabled. I'm now using PerfMon instead of AQtime; there are some PerfMon
statistics in a reply lower in this thread.

I'll answer your questions:
1. The same amount of work is being done, the same input data file contents,
the same executable being run, and results are the same.
2. Run times of 8 minutes on Windows XP and 35 minutes on Windows 2003 Server.
3. Microsoft Windows XP Professional version 2002 (5.1.2600 Service Pack 2
Build 2600), and Microsoft Windows Server 2003 Enterprise Edition (5.2.3790
Service Pack 2 Build 3790). We have had reports of the same problem at other
sites, I do not know what versions of the operating system they are using.
4. Context switch rate during one such program run:
. XP average 1,100 Context Switches/sec
. 2003 Server 230 Context Switches/sec
5. There is some disk I/O associated with the program, and network I/O is
almost entirely from the Remote Desktop Connection I use to connect to the
computers. These are the same in both cases, I am the only user on the
machines and this is the only program I run on them, apart from PerfMon, and
the operating system. the CPU is an average of just under 50% User Time (with
Hyperthreading) and just under 100% User Time (with HT disabled).
6. No difference with/without HT, single core or dual processor.
7. From the AQtime help: "A cache miss is an event that occurs when the CPU
is trying to read data from the cache, but this data is not in the cache.
Cache misses reduce the application performance because the CPU needs to
access the next-level cache or the main memory". AQtime is a very invasive
tool, and slows programs down by a factor of about 10, which may skew results
considerably. Although very useful for finding other bottlenecks, I am now
using PerfMon.

I would like to open a case support issue, and copy all these discussions
into it. I've answered your questions here is case someone else has any
suggestions.

All the best,
Eddie

""Jeffrey Tan[MSFT]"" wrote:

Hi Eddie,

Thanks for your feedback.

I have discussed this issue with the Windows kernel team. We believe this
issue is a bit complex to troubleshoot. Below are some questions regarding
your problem:

1. It¡¯s important to make sure the measurement is being done consistently
and correctly. Is the same amount of work being done in all cases? How is
elapsed time being measured? How is CPU usage being measured? Are both Oses
seeing the same number of CPUs?
2. What sorts of run-times are you talking about? Seconds? Minutes? Hours?
3. What OS revisions are you using?
4. What is the context switch rate for XP and Windows Server2003?
5. Is there disk/network I/O?
6. What happens if you turn off hyperthreading in BIOS? Boot with
numproc=1? Do you see a difference?
7. What do you mean by memory cache misses? Processor cache misses? File
cache misses? Page faults? Can you quantify the difference?

PS. if this issue is somewhat urgent, I would recommend you to contact
Microsoft CSS for a case support. This is a more efficient way of
troubleshoot this issue. If you wanted to contact Microsoft CSS, please
feel free to tell me, I will provide you the contact info.

Thanks.

Best regards,
Jeffrey Tan
Microsoft Online Community Support
==================================================
Get notification to my posts through email? Please refer to
http://msdn.microsoft.com/subscriptions/managednewsgroups/default.aspx#notif
ications.

Note: The MSDN Managed Newsgroup support offering is for non-urgent issues
where an initial response from the community or a Microsoft Support
Engineer within 1 business day is acceptable. Please note that each follow
up response may take approximately 2 business days as the support
professional working with you may need further investigation to reach the
most efficient resolution. The offering is not appropriate for situations
that require urgent, real-time or phone-based interactions or complex
project analysis and dump analysis issues. Issues of this nature are best
handled working with a dedicated Microsoft Support Engineer by contacting
Microsoft Customer Support Services (CSS) at
http://msdn.microsoft.com/subscriptions/support/default.aspx.
==================================================
This posting is provided "AS IS" with no warranties, and confers no rights.


.

Relevant Pages

  • Re: 2.6.16-rc6-mm2
    ... CPU: Trace cache: 12K uops, ... Calibrating delay using timer specific routine.. ... # ACPI Support ...
    (Linux-Kernel)
  • RE: dfsserv.exe using too much cpu
    ... I understand that you notice the dfsserv.exe consuming large CPU on your ... you contact Microsoft Product Support Services via telephone for a better ... solution and a dedicated Support Professional can assist with your request. ... This newsgroup only focuses on SBS technical issues. ...
    (microsoft.public.windows.server.sbs)
  • RE: Cannot change any control id in vs2005 web project
    ... existing issue of the Visual Studio 2005 IDE. ... And each web project will has a own sub cache directory under it. ... Microsoft CSS for a hotfix against this issue. ... Microsoft MSDN Online Support Lead ...
    (microsoft.public.vsnet.general)
  • RE: CPU and Services
    ... Thank you for posting to the SBS Newsgroup. ... much CPU, and you cannot start one service in the services console. ... we suggest that customers contact Microsoft Product Support Services ...
    (microsoft.public.windows.server.sbs)
  • RE: step in .NET makes vs2005 freeze when unmanaged code is enabled
    ... this should be a low CPU hang, which will not cause the 100% CPU. ... You may get Process Monitor from the link below(it also contains a video ... Microsoft Online Community Support ...
    (microsoft.public.vsnet.debugging)
Loading