can someone have idea on this perfmon result

15 Sep 2007 4:17 AM

fqiesbry

I have setup perfmon counter in our DB server. some of the counters
are Ave. disk read queue lenght, % processor time, batch request/sec
and other counters. the interval is every 15 sec. I saw some entries
that started some perf issue in our web-based application:

Here are the data:

Time: 15:23:15
Ave. disk read queue lenght = 0.129
% processor time = 0
batch request/sec = 322.4

Time: 15:23:37
Ave. disk read queue lenght = 0
% processor time = 82.66
batch request/sec = 2.5

Time: 15:23:52
Ave. disk read queue lenght = 0.01
% processor time = 0.727
batch request/sec = 10.14

then perfformance issue was reported in application side. The number
of connection increased to 300+, normal is 150 only.

What could be happen in 15:23:37 wherein the IO become less but CPU is
high, batch request become slow then gradually it increased? Any idea
is very appreciated.

Thanks.

15 Sep 2007 2:18 PM

Andrew J. Kelly

There is absolutely no way to tell from those counters alone. You need to
run a trace to see what commands are happening when these slowdowns occur.
One thing that may be a possibility is if some kind of maintenance or batch
operation was running that may have blocked the other users. That would
potentially reduce the batch requests per second since they ware all waiting
on the locks to go away.

--
Andrew J. Kelly SQL MVP
Solid Quality Mentors

Show quote

"fqiesbry" <fqies_***@yahoo.com> wrote in message
news:1189829834.865047.158920@g4g2000hsf.googlegroups.com...
>I have setup perfmon counter in our DB server. some of the counters
> are Ave. disk read queue lenght, % processor time, batch request/sec
> and other counters. the interval is every 15 sec. I saw some entries
> that started some perf issue in our web-based application:
>
> Here are the data:
>
> Time: 15:23:15
> Ave. disk read queue lenght = 0.129
> % processor time = 0
> batch request/sec = 322.4
>
> Time: 15:23:37
> Ave. disk read queue lenght = 0
> % processor time = 82.66
> batch request/sec = 2.5
>
> Time: 15:23:52
> Ave. disk read queue lenght = 0.01
> % processor time = 0.727
> batch request/sec = 10.14
>
> then perfformance issue was reported in application side. The number
> of connection increased to 300+, normal is 150 only.
>
> What could be happen in 15:23:37 wherein the IO become less but CPU is
> high, batch request become slow then gradually it increased? Any idea
> is very appreciated.
>
> Thanks.
>