Re: scaling performance vs number of CPUs

[Jay Pipes]

Well, here's my thoughts on your situation... If the threads will be examining the same set of record data, I don't think there's a whole lot of benefit to using multiple threads to do the SELECTs. IMHO, you're just going to run into the race conditions that you saw earlier. The only way that you can circumvent the next-key locking (really, just circumventing the row-level locking) would be to lock the table as a whole using LOCK TABLE my_table READ; before each SELECT statement. This would perhaps be slightly faster than going with the row-level locking in your situation because the threads are all examining the same range of records. Locking the table as a whole would certainly prevent the race conditions for the row-level locks (because other threads would simply wait for the table-level lock to be released). However, you may run into a situation where threads will timeout waiting in line for the table-lock to be released. Perhaps it's worth testing. Try it and post back your results...

Jay Pipes
Community Relations Manager, North America, MySQL Inc.

Got Cluster? http://www.mysql.com/cluster
Personal: http://jpipes.com



Edited 1 time(s). Last edit at 09/25/2005 07:58AM by Jay Pipes.