Apache1.3.12/PHP-4.0.0: Peformance on AIX (2:nd try)

From: Date: Thu, 08 Jun 2000 16:31:38 +0000
Subject: Apache1.3.12/PHP-4.0.0: Peformance on AIX (2:nd try)
Groups: php.general 
Request: Send a blank email to php-general+get-1312@lists.php.net to get a copy of this message
Hi all! I do need some hints to find the performance bottleneck of our Apache/PHP system running on an IBM H70. Here are some facts: Each IBM H70 have 1 GB memory, dual 340 MHz RS64-II processors with 4 MB L2 cache, 1 systemdisk, 3x18 GB disks in SW raid 0 for pages. During high traffic time, each H70 can get up to 200 simultaneous requests, but a that time they start to choke. And this is not what we expected when moving from a cheap Linux-solution to this expensive IBM-system. What we need is that each H70 should be able to handle 1000 simultaneous request with good performance. When I initially start the Apache server, I get around 15 httpd-processes with most of them consuming 4-5 MB of memory (at least the SIZE column says so when running 'top', and they are all sleeping - 'slp'). Once I let the external traffic in, the httpd-processes grows to between 9 and 12 MB each in size. They are however going back and forth between 'run' and 'slp' according to 'top', but they are still consuming these 9-12 MB even when sleeping ('slp'). If I stop the traffic to the Apache server, all httpd-processes goes to sleep, but they are still consuming these 9-12 megs of memory (again according to 'top'). But they are dying one after the other with a speed of 1-per-5-seconds (aprox.). Here are some important numbers from our httpd.conf: Timeout 300 KeepAlive On MaxKeepAliveRequests 1000 KeepAliveTimeout 15 MinSpareServers 10 MaxSpareServers 25 StartServers 10 MaxClients 512 MaxRequestsPerChild 10000 ExtendedStatus Off I hope someone can help me with what to do next to find the problem? Or is it simply that we bought a to week system? Thanks in advance! /Per, Sweden

« previous php.general (#1312) next »