On Friday 31 October 2008 08:07:08 Sam Mason wrote: > On Thu, Oct 30, 2008 at 02:28:38PM -0700, Alan Hodgson wrote: > > On Thursday 30 October 2008, Joao Ferreira > > <jmcferreira@xxxxxxxxxxxxxxxxxx> > > > > wrote: > > > well..... see for yourself... (360 RAM , 524 SWAP) that's what it is... > > > it supposed to be somewhat an embedded product... > > > > Clearly your hardware is your speed limitation. If you're swapping at > > all, anything running on the machine is going to be slow. > > The vmstat output only showed the odd block going in and out; but > performance is only really going to suffer when it's thrashing. If the > swap in number stays in the double digits for a reasonable amount of > time then you should probably look at what's causing it. Giving memory > back to the system to use for caching the file system can be good, lots > of shared memory can also be good. > well, i think he needs to cut back on the work mem, but i think he might want to give a little more to wal buffers. > Building indexes takes time and IO bandwidth, maybe you could look at > building less of them? I'd be tempted to pull the import script apart > into its constituent parts, i.e. the initial data load, and then all the > constraints/index builds separately. Then run through executing them by > hand and see what you can change to make things more efficient. > It would be good to know where and when his bottlenecks are... ie. i could see him being i/o, memory, or cpu bottlenecked depending on where he is in the restore process. -- Robert Treat Conjecture: http://www.xzilla.net Consulting: http://www.omniti.com -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
