On 21/08/2013 1:31 a.m., inittab wrote:
Hello, Thanks for the suggestions. I've moved up to squid version 3.3.5, changed the raid5 into a raid0, tweaked the value of cache_dir to 100000, moved the acl manager lines, removed hierarchy_stoplist, and enabled memory_pools. I have also added RPS monitoring to our cacti instance so i can get a better idea when I give this a shot again. When running multiple processes of squid to deal with multiple (slow) cores, how many processes do you recommend running? 1 per core or less? I currently have 10 processes setup on 12 cores but do not know if this is the correct way to go about it.
Yes it is. Squid can possibly use the whole CPU core cycles, but each worker is a single threaded process so it cannot make use of more than one core. If you are using rock cache_dir they use a separate disker proces to manage teh disk I/O so it may be worthwhile allocating a core to that process as well as the main worker(s).
Amos
