Re: Excessive Paging

Hi Ray:

Without knowing what's running on the machine, all those statistics are just so many numbers.

Here's what matters: What applications are running and how much resources are they using?

There's no mystery behind slow performance on a machine without hardware problems. It's simply an imbalance of running applications vs. resources available to run them (principally memory). The solution is equally simple: Add resources or remove applications that use too much of them.

Earl Grey

Ray Manning wrote:
I'm running an XP SP2 machine that has experienced an incredible
decrease in performance over the past couple months and I'm trying to
determine why. At first I thought the hard drive might be at fault
since I was seening It would appear to me that the machine is
needlessly paging based on the PerfMon.exe Pages/sec counter and the
taskman performance numbers. Can someone verify I'm on the right track
and possibly offer solutions?

Here's what I'm seeing:

Perfmon.exe (Pages/Sec) averages of 30-60 with a scale of 1

TaskMan: Performance

Physical Memory (K)

Total : 392,496
Available: 170,000
System Cache 240,000


Commit Charge(K)

Total 158,568
Limit 550,000
Peak 170,000

CPU usage averages about 5-20% but of course spikes at times.


What I don't understand, and perhaps it's my interpretation of the
perfmon counter, is why I have such a high page/sec rate when my peak
commit charge isn't even half of the physical memory. Shouldn't my
commit charge be much higher before paging starts hitting 40+
pages/sec? Is page size 4k? 160k/second when I have a couple hundred
megs of physical memory available?

I did not look at these numbers before I started having performance
problems (long boot times, 5 minutes to log on, etc) so perhaps these
are not out of line, but if not, can someone explain to me what I'm
missing in my interpretation?

Thanks,
Ray

.

Relevant Pages

  • Re: Please explain "Always Running" mentality of Windows CE
    ... What event should I look for in .NETCF for an application de-activating? ... > because the file was currently opened by the applications. ... > release all opened resources and save its state. ... > the app is that it is faster, because all the initialization is already ...
    (microsoft.public.pocketpc.developer)
  • Re: [RFC][0/3] Virtual address space control for cgroups (v2)
    ... I thought I addressed some of those by adding a separate config option. ... > Trying to account/control physical memory or swap usage via virtual ... > swap space usage without breaking a large class of apps, ... As far as applications failing gracefully is concerned, ...
    (Linux-Kernel)
  • Re: Active/Passive to Active/Active
    ... A/A means separate resource groups running on separate nodes, ... spare available resources to handle the failover of one node to another ... You should NOT install applications in the Cluster Group on ... The articles you read are wrong, MSCS does NOT do loadbalancing. ...
    (microsoft.public.windows.server.clustering)
  • Re: [RFC][0/3] Virtual address space control for cgroups (v2)
    ... > Trying to account/control physical memory or swap usage via virtual ... > swap space usage without breaking a large class of apps, ... I find it hard to understand why applications would use sparse ...
    (Linux-Kernel)
  • Re: Please explain "Always Running" mentality of Windows CE
    ... I think well-programmed applications should close all the resources, ... de-activates it and hides it. ... The only difference between re-activating a de-activated app and starting ...
    (microsoft.public.pocketpc.developer)