Re: Overwhelming a Web Application Server

["Walden H. Leverich" <WaldenL@xxxxxxxxxxxxxxx>]

> But if you're willing to indulge in a somewhat hypothetical
discussion, what 
> specific measurements would you look at?

Perfmon is your friend (but I'm a windows guy). In all likelihood CPU
isn't the problem, most machines have more CPU than they know what to do
with. Still, I'd start w/perfmon on the web, app and DB servers. I'd
measure:

CPU (doesn't hurt to check), 
disk IO
disk queue
network io
network queue
db cache hit ratio
db connection counts
memory usage
page response times
page requests / second

All queues should ideally be 0, > 1 is a problem. 
Disk IOs should be < 15ms (preferable _way_ less). >25ms is horrid. 
DB cache should be _high_ 90s, like 99.something%.
page response times depends on SLA, but we look for well < 1s.

I'd also look at page payload sizes. the server performance is only one
part of the performance equation, and often a very small part. Server
roundtrips and page render time can often far outweigh the HTML
generation time.

These are of course general rules of thumb, there are always exceptions
and it's certainly situational.

-Walden