Re: PASE experts wanted

Joe,

It could be that my assumptions were wrong. I was basically under a DOS attack yesterday and the log was recording 12-15 hits every minute. The actual hit rate may have been higher and I was only seeing successful requests being logged. I am not familiar enough to know what each "trackback" request was triggering in the PHP app. There was DB I/O and disk I/O because the log, sessions and the database were being updated but I haven't actually reviewed the PHP code to see what else might have been happening.

I am not a performance guru by any stretch and perhaps looking at WRKACTJOB is the wrong way to go about determining "load". Here is a screen shot: http://www.petesworkshop.com/i/wrkactjob.jpg . All other jobs listed below this shot show 0 CPU usage. When I hit the site now, there is no change in overall CPU utilization so perhaps yesterday's events were anomaly. In any case, the http filtering stopped whatever events were sapping the CPU.

Pete

Joe Pluta wrote:

Hmm. 5% utilization on an empty box seem very excessive, in my opinion. I'll fire stuff up on my 270 this week and double check, but I typically never went over 2% on that box empty, either. And one hit per 3-4 seconds shouldn't make your system sneeze, much less fire up a bunch of jobs taking 5-8%. I thought PASE and/or PHP might be a bit of a hog, but I never dreamed it would be that bad.

And people complain about Java.

Joe

Joe,

I have a Websphere admin instance. A PHP instance running a weblog. Apache front ending a Ruby on Rails application (in PASE). And a JRuby application (in PASE).

Right now it is at 5.6% total utilization with nobody knocking on the door (that I know of).

Earlier, when I was getting hit every 3-4 seconds with a trackback request processed by the PHP app and written to the DB each PHP instance was showing 5-8% CPU and there are 25 ZENDCOREAP jobs running, presumably to service requests.

This 270 is "littler than yours" (never thought I'd boast about that!). 150 CPW. Hence the higher CPU utilization AFAIK.

Pete


Joe Pluta wrote:

Pete Helgren wrote:

Adding some filters to my firewall took care of it and now the CPU is idling at 10%.

That's interesting. I've got three complete WebSphere instances running on my little model 520 (one for my website, one for development and one for testing) and my CPU idles well under two percent. Heck, I just pummeled my website and I'm still under two percent. What the heck requires 5-8% per instance and 10% idle?

Joe