Re: ASP.Net Performance Problems

From: Neal (Neal_at_discussions.microsoft.com)
Date: 12/14/04 

Date: Tue, 14 Dec 2004 00:07:03 -0800

Hi,
I just received the results etc
They are

The stress testing was conducted over a week long period using the Microsoft
Web Application Stress 1.1 tool. During the testing I have monitored both
the webserver (DEVWEB) and the database server (DEVDB).

Web server :1 x 350 MHz Pentium...{May Explain a lot}
512 Mb RAM
28GB Hard disk
Windows 2000 Server SP4IIS 5.0

10 Concurrent Users exercising all urls,

Web Server

Disk I/O - OK
Memory utilisation - OK
CPU - overloaded

DB Server

Disk I/O - OK
Memory utilisation - OK
CPU - OK

The tests that we performed were directly monitored on each server.
Supporting data is attached.

We found that 10 concurrent users maxed out the
web servers CPU and pushed the processor queue length way beyond its
acceptable tolerance. It should not exceed 2, whereas the tests we
performed logged it at circa > 20.

The performance test data shows that the ASPNET_WP application CPU
utilisation hovers above the 75% critical level if the stress level (number
of Threads) is increased to 10 or more. The number of Threads equates to the
number of concurrent connections. The total CPU utilisation borders on 100%
for the duration of the test. The processor queue length averages about 23,
071 indicating that a number of threads are being queued as the processor
cannot coupe with the load. This indicates that there is a CPU bottleneck
for the current server configuration.

The data also shows that disk IO is not a problem. The tests showed that
the performance counters are well within the recommended values. The % disk
idle time averages about 99% which indicates very little disk reads. Memory
usage does not seem to be a problem at all. The % committed bytes in use
does not exceed the 25% mark for the full duration of the test.

This will be tested on a Dual Pentium Processor (the Varsity's Production
WebServer), next, but .....!!

So seems the number of threads created, used pooled
instantiated/de-instantiated may be the problem...Just guessing, I know squat
about this.

thanks
Neal

"Sam Santiago" wrote:

> Which CPU was maxed out - the DB or the Web server? You should probably run
> a profiling tool during the test to determine which class is the culprit. I
> personally like a tool from AutomatedQA called AQTime4 -
> http://www.automatedqa.com. Microsoft offers a freebie that is not quite
> as user friendly, but effective once you figure it out:
> http://msdn.microsoft.com/netframework/downloads/default.aspx.
>
> Thanks,
>
> Sam
>
> --
> _______________________________
> Sam Santiago
> ssantiago@n0spam-SoftiTechture.com
> http://www.SoftiTechture.com
> _______________________________
> "Neal" <Neal@discussions.microsoft.com> wrote in message
> news:FDA668A3-5396-4CC1-A47C-498003A3DC4D@microsoft.com...
> > Hi All
> >
> > An app I wrote (ASP.NET) against SQL Server was Load Tested by our IS
> > division, and they came back with a max of 10 concurrent users before the
> > Server's CPU was overloaded.
> > Running ACT against the same app on the same Server, I was getting 1746
> > concurrent users before a time out occurred. (using the ACT graphs showing
> > RPS and TLB )
> >
> > I am using StoredProcs, and the SQL server optimised providers (in code),
> > populating datasets, with some caching, to Grids and labels and DropDowns,
> > ..nothing unusual or rocket science.
> > The datasets are also optimised (kept as small as possible)
> >
> > Is this possible , surely not with .Net ??, could there be an explanation
> > in
> > the deployment, IIS or in the app??
> >
> > Thanks
> > Neal
> >
> >
> > --
> > Neal Rogers
> > University of Cape Town
>
>
>

Relevant Pages

  • Re: ASP.Net Performance Problems
    ... for the high CPU usage. ... > the webserver and the database server. ... > Disk I/O - OK ... >> a profiling tool during the test to determine which class is the culprit. ...
    (microsoft.public.dotnet.distributed_apps)
  • Re: [opensuse] Best CPU for OSS 10.3 Small Home Network Server
    ... server" ... ... CPU's are thousands of times faster than the fastest disk ... Don't buy into the propaganda that CPU speed is the primary ... CPU intensive task, and printing is a task which CAN BE ...
    (SuSE)
  • Re: We quadrupled hardware power and reduced performance
    ... Seems to be CPU where the bottleneck is, ... > CPU, Memory, Disk i/o, Index information, long running queries etc. ... >> We did several tests with new box serving as server to other boxes. ...
    (microsoft.public.sqlserver.server)
  • Re: [opensuse] Best CPU for OSS 10.3 Small Home Network Server
    ... CPU's are thousands of times faster than the fastest disk ... Don't buy into the propaganda that CPU speed is the primary ... determinant of system performance...and that a "server" ... I have enough memory to put 768 MB RAM in it. ...
    (SuSE)
  • hai...find me a solution in M5000 server
    ... series server. ... B B B B B i want to do install one more solaris Over there. ... Solaris cannot see all drives on Areca RAID controller ... popped it (a single disk at this point) into machine1. ...
    (SunManagers)