Re: Linux and mm session handler -- does it really handle high load?

Found the problem. It's actually in output layer, not the patch. Just make sure you don't output any before starting session/sending headers. PHP fails to complain about it under my system and it caused segfaults. Here is my simple performance test reuslt [root@dev root]# ab -c 500 -n 5000 http://tmp-dev.localnet/~yohgaki/test/sess_save_handler/mm-handler.php This is ApacheBench, Version 1.3d <$Revision: 1.58 $> apache-1.3 Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/ Copyright (c) 1998-2001 The Apache Group, http://www.apache.org/ Benchmarking tmp-dev.localnet (be patient) Completed 500 requests Completed 1000 requests Completed 1500 requests Completed 2000 requests Completed 2500 requests Completed 3000 requests Completed 3500 requests Completed 4000 requests Completed 4500 requests Finished 5000 requests

Server Software:        Apache/1.3.23
Server Hostname:        tmp-dev.localnet
Server Port:            80

Document Path:          /~yohgaki/test/sess_save_handler/mm-handler.php
Document Length:        30 bytes

Concurrency Level:      500

Time taken for tests: 37.948 seconds

Complete requests:      5000
Failed requests:        0
Broken pipe errors:     0
Total transferred:      1923375 bytes
HTML transferred:       153870 bytes
Requests per second:    131.76 [#/sec] (mean)
Time per request:       3794.80 [ms] (mean)
Time per request:       7.59 [ms] (mean, across all concurrent requests)
Transfer rate:          50.68 [Kbytes/sec] received

Connnection Times (ms)

              min  mean[+/-sd] median   max
Connect:        3   539  656.8    432  9593

Processing: 496 1612 2533.7 1323 33401

Waiting:      411  1611 2533.7   1323 33402
Total:        496  2150 2886.2   1872 37037

Percentage of the requests served within a certain time (ms) 50% 1872 66% 1961 75% 2043 80% 2072 90% 2290 95% 2413 98% 5657 99% 11294 100% 37037 (last request) [root@dev root]# BTW, this Linux is running with 433Mhz Celeron PC. If you find problems, let me know. I might be able to fix it before commit. -- Yasuo Ohgaki Yasuo Ohgaki wrote:

Looks like I've done something wrong. Older patch performs well but not this one. Don't even try the patch. -- Yasuo Ohgaki Rick Morris wrote:

I have tried mm as a session handler three times, on three different Linux boxes. Each one was a server under heavy load, running many concurrent sessions, so I was hoping mm would solve some performance issues, but have been disappointed each time.