Application & Session Cache Performance Problem

[roadrunner]

Hi Guys,

After some initial tests using loadrunner it looks like just storing the
Application("Items") array into a local array and then accessing that gives a
significant improvement in performance. It seems to run in a tenth of the
time that the original code took no matter how many simultaneous requests
there are.

Plotting response time vs. Number simultaneous users gives a linear graph
for both versions of the code. This points to just poor performance rather
than a bottleneck being the main problem of the original code, surely? E.g.
It was running slow due to the cpu taken to copy the array rather than
threads blocking one another. If there was a bottleneck we would see a point
on the graph where response times suddenly escalate.

Thanks

 

[Anthony Jones]

Hi Guys,

After some initial tests using loadrunner it looks like just storing the
Application("Items") array into a local array and then accessing that gives a
significant improvement in performance. It seems to run in a tenth of the
time that the original code took no matter how many simultaneous requests
there are.

Plotting response time vs. Number simultaneous users gives a linear graph
for both versions of the code. This points to just poor performance rather
than a bottleneck being the main problem of the original code, surely? E.g.
It was running slow due to the cpu taken to copy the array rather than
threads blocking one another. If there was a bottleneck we would see a point
on the graph where response times suddenly escalate.

How many CPUs do you have

A runnable thread is still effectively blocked if it can't find a processor
to run on.

If you had more than one and all were busy in the original incarnation of
the site that would indicate that ASP was allowing multiple reads of the
array. That would be good know if not particularly helpful here.

Anthony.

 

[roadrunner]

Good point. In live we have multi cpu machines but the test machine I have
available is only a single cpu. Guess it's like comparing apples to pears..

It does seem to greatly improve performance, however, so I think we will
have to try this in live to really see how much of an improvement it makes.
Not ideal but c'est la vie..