my virtual host has Intel(R) Xeon(R) CPU E5-2620 @ 2.00GHz 2processor, 2GB memory.apache server.
each participant take the cpu usage higher than 10% even run a simple survey. which means my virtual host can only afford 10 participant at the same time.
why??
i test joomla, drupal with more than 500M mysql database, search only take 4% cpu usage.
because the yii framework? or coding problem?

It is a virtual server so you will never really know what you get, they might have imposed some restrictions you don't know about.

I'd also strongly recommend to use a PHP runtime cache (like eaccelerator.net/ ), that will greatly improve performance.

this virtual host in my server, i think the problem is coding problem.
files larger than 50M for install, i just guess it can be lighter.i think use zend can find out where is the neck.

iscar wrote: each participant take the cpu usage higher than 10% even run a simple survey. which means my virtual host can only afford 10 participant at the same time.

How did you measure that you only can serve 10 concurrent users with your setup.

What OS is used? Windows? Linux?
How are Apache and PHP used together? CGI,DSO,FastCGI PHP-FPM oder suPHP?

A problem with Limesurvey is the big session file which is written for every new user starting the survey. If the session files are written to tmp and tmp is setup as a separate partition the free space is used up quickly if the session timeout is high. That might explain some performance issued when many respondents start the survey at the same time, but not during the interview.

Is this regarding the 1000+questions questionnaire here?
www.limesurvey.org/en/community-services...500-questions-survey

the server is centos system, apache prefork with php module.
we monitor the server when user answer the survey, it's the 1question survey, only for test.which means even the most simple question, ls use 10%+ cpu. as the much complex cms software, use only 4% cpu.
the session file use memory only, but how to explain use 10%+ cpu even under most simply survey?

I don't think that this is a normal behavior.

Maybe there is something wrong (installation / configuration / survey).

Let's see if some of the developers can shed some light.

iscar wrote: the server is centos system, apache prefork with php module.
we monitor the server when user answer the survey, it's the 1question survey, only for test.which means

What CentOS? x64,x32 5,6?
What Apache version?
What PHP Handler?
What PHP Cache Handling?

How did you measure that you only can serve 10 concurrent users with your setup?
Which depends not on CPU alone but on memory and bandwidth as well.
I don't see any sense in looking at the cpu usage during the init process.
Limesurvey uses a demanding init process when creating the sessionfile.

I wonder how you did enforce that no session file is created via php on the storage?
You said that session files are memory only? Really? What is the session path in php.ini?
Is that pointing to a ram disk path?

Even when using Apache 2.2 with a SuPHP (which is the most CPU demanding PHP-handler) Limesurvey can handle more than 10 concurrent users (even on a virtual server in a priceclass < 50 USD per Month).

If you optimize the code of Limesurvey to use less CPU cycles, go on. Still room for improvement.

we tested ls and other cms in different version centos,even use a new server to test. the TYPICAL LAMP environment is :centos 6.4+apache 2.2+mod_cache.so+mod_disk_cache.so+libphp5.so
THE RESULT ARE SAME:ls use more 200%+ cpu than joomla, drupal,even the most simple survey.
then we need to find out the reason as to improvement.

i am not sure if ls has performance test before release, i monitor the server performance in KVM platform.OR use top command OR ps aux | awk 'BEGIN {CPU=0;MEM=0} /httpd/ {CPU = CPU + $3 ; MEM = MEM + $4.it's easy to monitor the cpu usage in linux.

this problem is not related to the server,i think we need to find out the coding problem, where eat lots of cpu.