Hardware and performance

Am by no means an expert on jmeter testing, but would think that 100 users at one time might be sufficient for one course depending upon the course content and how 'class is conducted'.  Class to be held face to face with 100 students using whatever device and hitting Moodle at same time (like a HS class with 100 students - large class, BTW  That legal?)?  OR is the class like that of 'typical' higher ed ... students access whenever?

Bottle necks could be present in MySQL, PHP,  or Apache (assuming typical install).  And bottle neck could be in Virtualized environment (if running under a Virt) or even in networking (behind firewall).

Done any tweaks to MySQL (assuming local host - ie, DB server and DB for Moodle on same server as the web server + Moodle code/app)?   If running Linux, for example, one could also acquire and run mysqltuner (or other such script) to check the DB server settings.   How about Apache server itself?  One could do some simple 'apache benchmark' (on linux man ab) test to see if tweaking apache settings wouldn't improve jmeter responses.

When running jmeter test did you check 'top' to see if swap space was being used (that has to do with memory management)?   If so, did you also investigate what was using swap space (top with FP)?

Did you check apache error logs to see if there any clues there?

Probably should have asked first ... what does 'not reponding' look like?   That sounds like a dumb question but not really - browser might not get page rendered as quickly as before and it actually might time out ... but services on the server (apache/mysqld) are still up and running.   PHP settings/tweaks for how much memory a script can consume + how long a script is given to run (30 seconds is default).

My 2 cents! ;)

'spirit of sharing', Ken

Have you been through http://docs.moodle.org/27/en/Performance_recommendations ?