On Fri, Jan 23, 2009 at 12:39 PM, Brian Aker <br...@tangent.org> wrote: > Hi! > > On Jan 24, 2009, at 2:33 AM, MARK CALLAGHAN wrote: > >> The results in the bug are measured on a 4 core server. I have results >> that are at least as bad from a 16 core server. When pool-of-threads >> is used _every_ command sent to the server for every connection goes >> through a pthread_mutex lock/unlock on LOCK_event_loop. > > One thing I have found is to set the number of threads at or below the > number of cores that are available.
And that will kill performance for anything in which connections block in the server or hang the server when all threads are blocked. > > Cheers, > -Brian > > -- > _______________________________________________________ > Brian "Krow" Aker, brian at tangent.org > Seattle, Washington > http://krow.net/ <-- Me > http://tangent.org/ <-- Software > _______________________________________________________ > You can't grep a dead tree. > > > > -- Mark Callaghan mdcal...@gmail.com _______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : drizzle-discuss@lists.launchpad.net Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
