Comments on: Effects on Performance & Server Resources when using Apache 2.0 mod_status http://www.philchen.com/2008/06/02/apache-20-mod_status-effects-on-performance-server-resources Phil Chen Sun, 22 Jan 2012 06:08:55 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: Oleg Vishnepolsky http://www.philchen.com/2008/06/02/apache-20-mod_status-effects-on-performance-server-resources/comment-page-1#comment-2093 Oleg Vishnepolsky Tue, 14 Apr 2009 17:51:02 +0000 http://www.philchen.com/2008/06/02/apache-20-mod_status-effects-on-performance-server-resources/#comment-2093 Is the period over each the averages for requests per second configurable ? We need to know how many requests we are getting per second closer to real time. I would assume that this is a common request. Is the period over each the averages for requests per second configurable ? We need to know how many requests we are getting per second closer to real time. I would assume that this is a common request.

]]> By: Phil Chen http://www.philchen.com/2008/06/02/apache-20-mod_status-effects-on-performance-server-resources/comment-page-1#comment-1759 Phil Chen Wed, 12 Nov 2008 02:26:49 +0000 http://www.philchen.com/2008/06/02/apache-20-mod_status-effects-on-performance-server-resources/#comment-1759 Yep definitely over the period of the uptime. Regarding: "In contrast, a chart of BusyWorkers is much more spiky, sharply peaking at 09.00h and diving at lunch-time and generally behaving much more like I know the traffic to be." I find that when you run KeepAlive On that will effect the spiky BusyWorkers also. Via the KeepAliveTimeout and Min / Max Servers. Yep definitely over the period of the uptime.

Regarding:

“In contrast, a chart of BusyWorkers is much more spiky, sharply peaking at 09.00h and diving at lunch-time and generally behaving much more like I know the traffic to be.”

I find that when you run KeepAlive On that will effect the spiky BusyWorkers also. Via the KeepAliveTimeout and Min / Max Servers.

]]> By: Owen Boyle http://www.philchen.com/2008/06/02/apache-20-mod_status-effects-on-performance-server-resources/comment-page-1#comment-1758 Owen Boyle Tue, 11 Nov 2008 10:06:02 +0000 http://www.philchen.com/2008/06/02/apache-20-mod_status-effects-on-performance-server-resources/#comment-1758 Me again... I just RTFC'd and, would you believe it, the averages are calculated over the up-time! So ReqPerSec is not a rolling average at all, it's a life-time average. So no wonder it smooths out; if the server stays up long enough, it'll end up a flat line. Not really what I was looking for - maybe it's time to Hack-TFC :-) Me again… I just RTFC’d and, would you believe it, the averages are calculated over the up-time! So ReqPerSec is not a rolling average at all, it’s a life-time average. So no wonder it smooths out; if the server stays up long enough, it’ll end up a flat line.

Not really what I was looking for – maybe it’s time to Hack-TFC :-)

]]> By: Owen Boyle http://www.philchen.com/2008/06/02/apache-20-mod_status-effects-on-performance-server-resources/comment-page-1#comment-1757 Owen Boyle Tue, 11 Nov 2008 10:00:13 +0000 http://www.philchen.com/2008/06/02/apache-20-mod_status-effects-on-performance-server-resources/#comment-1757 Interesting analysis. I have also recently started using mod_status and find it makes no difference to the server load - even in extended mode. So it seems that old chestnut is a bit of a red herring... I came across your blog because I was trying to find out how mod_status calculates its averages - they seem to be rolling averages over a very long period (24h?) so that they change only very slowly. If I look at a chart of ReqPerSec against time, it swoops very gently up and down over the day, peaking at 50 and dipping to 20 over a period of about 12h. In contrast, a chart of BusyWorkers is much more spiky, sharply peaking at 09.00h and diving at lunch-time and generally behaving much more like I know the traffic to be. Do you have any insight into how the averages are worked out and - are they configurable? (I guess, I could go read the code... :-) Interesting analysis. I have also recently started using mod_status and find it makes no difference to the server load – even in extended mode. So it seems that old chestnut is a bit of a red herring…

I came across your blog because I was trying to find out how mod_status calculates its averages – they seem to be rolling averages over a very long period (24h?) so that they change only very slowly. If I look at a chart of ReqPerSec against time, it swoops very gently up and down over the day, peaking at 50 and dipping to 20 over a period of about 12h. In contrast, a chart of BusyWorkers is much more spiky, sharply peaking at 09.00h and diving at lunch-time and generally behaving much more like I know the traffic to be.

Do you have any insight into how the averages are worked out and – are they configurable?

(I guess, I could go read the code… :-)

]]>