Home All Groups Group Topic Archive Search About
Author
29 Jul 2006 1:00 AM
Agi
Hello, theres,
  I set up a simple environment, 2 machines, one for sql 2000 EE(2 HT
CPUs, 2G RAM) , one for IIS + .net 2.0 (2 CPUs, 1G Ram)
  In sql server,I create a simple table UserPwd (20000 records) to
store userid/password then write a simple procedure to do login, I set
userid as primary key. Then using Microsoft Stress Tool to stress my
web application, I got 150 RPS. why only 150 ?
( the login.aspx just retuern ok if login successfully, connection
pooling is also on)

I monitor the SQL host, CPU utilization under 5, the AP server CPU also
under 30.

Any idea to increase the RPS ? any related article/resources?

Regards,
Agi Chen

Author
29 Jul 2006 1:21 PM
Dan Guzman
> I monitor the SQL host, CPU utilization under 5, the AP server CPU also
> under 30.

Are you running a single thread with the test tool?  It doesn't look like
SQL Server is the bottleneck.  Based on 30% CPU, it seems to me that the
application server CPU might be the limiting factor with this test.  I would
expect you could at least double or triple RPS by adding an more test
threads until you run out of app server CPU headroom.

--
Hope this helps.

Dan Guzman
SQL Server MVP

Show quote
"Agi" <agic***@gmail.com> wrote in message
news:1154134809.893931.77580@h48g2000cwc.googlegroups.com...
> Hello, theres,
>  I set up a simple environment, 2 machines, one for sql 2000 EE(2 HT
> CPUs, 2G RAM) , one for IIS + .net 2.0 (2 CPUs, 1G Ram)
>  In sql server,I create a simple table UserPwd (20000 records) to
> store userid/password then write a simple procedure to do login, I set
> userid as primary key. Then using Microsoft Stress Tool to stress my
> web application, I got 150 RPS. why only 150 ?
> ( the login.aspx just retuern ok if login successfully, connection
> pooling is also on)
>
> I monitor the SQL host, CPU utilization under 5, the AP server CPU also
> under 30.
>
> Any idea to increase the RPS ? any related article/resources?
>
> Regards,
> Agi Chen
>
Author
31 Jul 2006 5:00 AM
Agi
Dan ,
     I used more then 8 threads, finally my colleague told me to
disable the session state, then performance boost.

Regards
Agi

Dan Guzman 寫道:

Show quote
> > I monitor the SQL host, CPU utilization under 5, the AP server CPU also
> > under 30.
>
> Are you running a single thread with the test tool?  It doesn't look like
> SQL Server is the bottleneck.  Based on 30% CPU, it seems to me that the
> application server CPU might be the limiting factor with this test.  I would
> expect you could at least double or triple RPS by adding an more test
> threads until you run out of app server CPU headroom.
>
> --
> Hope this helps.
>
> Dan Guzman
> SQL Server MVP
>
> "Agi" <agic***@gmail.com> wrote in message
> news:1154134809.893931.77580@h48g2000cwc.googlegroups.com...
> > Hello, theres,
> >  I set up a simple environment, 2 machines, one for sql 2000 EE(2 HT
> > CPUs, 2G RAM) , one for IIS + .net 2.0 (2 CPUs, 1G Ram)
> >  In sql server,I create a simple table UserPwd (20000 records) to
> > store userid/password then write a simple procedure to do login, I set
> > userid as primary key. Then using Microsoft Stress Tool to stress my
> > web application, I got 150 RPS. why only 150 ?
> > ( the login.aspx just retuern ok if login successfully, connection
> > pooling is also on)
> >
> > I monitor the SQL host, CPU utilization under 5, the AP server CPU also
> > under 30.
> >
> > Any idea to increase the RPS ? any related article/resources?
> >
> > Regards,
> > Agi Chen
> >
Author
31 Jul 2006 12:30 PM
Dan Guzman
> finally my colleague told me to
> disable the session state, then performance boost.

This is good advice, unless of course the application requires session state
to be saved ;-)  If you save state to a database, is especially important
that the log be on a separate disk and configured for high-performance
writes (e.g. RAID 1, write-caching controller).

Before stress testing with many threads, I usually perform a performance
tests with a single thread in order to identify performance issues with
configuration, app code and database.  Once all is tuned, I then run stress
tests to ensure scalable performance.

--
Hope this helps.

Dan Guzman
SQL Server MVP

"Agi" <agic***@gmail.com> wrote in message
news:1154322013.748049.102300@m79g2000cwm.googlegroups.com...
Dan ,
     I used more then 8 threads, finally my colleague told me to
disable the session state, then performance boost.

Regards
Agi

Dan Guzman ??:

Show quote
> > I monitor the SQL host, CPU utilization under 5, the AP server CPU also
> > under 30.
>
> Are you running a single thread with the test tool?  It doesn't look like
> SQL Server is the bottleneck.  Based on 30% CPU, it seems to me that the
> application server CPU might be the limiting factor with this test.  I
> would
> expect you could at least double or triple RPS by adding an more test
> threads until you run out of app server CPU headroom.
>
> --
> Hope this helps.
>
> Dan Guzman
> SQL Server MVP
>
> "Agi" <agic***@gmail.com> wrote in message
> news:1154134809.893931.77580@h48g2000cwc.googlegroups.com...
> > Hello, theres,
> >  I set up a simple environment, 2 machines, one for sql 2000 EE(2 HT
> > CPUs, 2G RAM) , one for IIS + .net 2.0 (2 CPUs, 1G Ram)
> >  In sql server,I create a simple table UserPwd (20000 records) to
> > store userid/password then write a simple procedure to do login, I set
> > userid as primary key. Then using Microsoft Stress Tool to stress my
> > web application, I got 150 RPS. why only 150 ?
> > ( the login.aspx just retuern ok if login successfully, connection
> > pooling is also on)
> >
> > I monitor the SQL host, CPU utilization under 5, the AP server CPU also
> > under 30.
> >
> > Any idea to increase the RPS ? any related article/resources?
> >
> > Regards,
> > Agi Chen
> >

AddThis Social Bookmark Button