Dear all, I'm running MySQL 5.0.82 with a 16GB database, on a dedicated machine with 16 cores (4 x 4) and 64 GB of memory. It has 100 tables, of which 2 have been recently converted to InnoDB to alleviate table locks which were causing problems, and the rest are MyISAM.
Observing the system under load again, it seems that at a certain point the increase in load is no longer linear with the increase in concurrent users. At that point, the system starts doing more context switching (from ~3000, to ~160,000). Is that enough to conclude that the problem here is thread thrashing? What can I do in MySQL, and more specifically, MyISAM in order to possibly control this? Do you think the situation can be improved by switching to InnoDB? I've been reading that MySQL performance on SMP has been improved (or is in the process of), however it's unclear to me which improvements made it to 5.0.x, if any, and whether the improvements are for InnoDB only or MyISAM too. Any suggestions or pointers are very welcome. Thank you. Kind Regards, -- Amr Mostafa
