Re: [PERFORM] Interesting performance behaviour

2004-09-10 Thread Tom Lane
Joey Smith <[EMAIL PROTECTED]> writes: > EXPLAIN ANALYZE SELECT message_idnr FROM messages WHERE mailbox_idnr > = 1746::bigint AND status<2::smallint AND seen_flag = 0 AND unique_id > != '' ORDER BY message_idnr ASC LIMIT 1; > QUERY P

Re: [PERFORM] Interesting performance behaviour

2004-09-10 Thread Joey Smith
> > shared_buffers = 1000 > > sort_mem = 1024 > > effective_cache_size = 1000 > > effective_cache_size should be much higher, like 3/4 of your available RAM. > This is probably the essence of your planner problem; the planner thinks you > have no RAM. I set effective_cache_size to 64000 on a mach

Re: [PERFORM] Interesting performance behaviour

2004-09-10 Thread Josh Berkus
Joey, > shared_buffers = 1000 > sort_mem = 1024 > effective_cache_size = 1000 effective_cache_size should be much higher, like 3/4 of your available RAM. This is probably the essence of your planner problem; the planner thinks you have no RAM. > I see a similar speedup (and change in query pl