From: | Robert Haas <robertmhaas(at)gmail(dot)com> |
---|---|
To: | Chris Dunn <chris(dot)dunn(at)bigredsky(dot)com> |
Cc: | "pgsql-performance(at)postgresql(dot)org" <pgsql-performance(at)postgresql(dot)org> |
Subject: | Re: FW: Performance 8.4.0 |
Date: | 2009-08-03 02:16:50 |
Message-ID: | 603c8f070908021916r48ebab0ya43e556add3e6a00@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
On Sun, Aug 2, 2009 at 10:04 PM, Chris Dunn<chris(dot)dunn(at)bigredsky(dot)com> wrote:
> The database is 8gb currently. Use to be a lot bigger but we removed all large objects out and developed a file server storage for it, and using default page costs for 8.4, I did have it changed in 8.1.4
You might want to play with lowering them. The default page costs
make page accesses expensive relative to per-tuple operations, which
is appropriate if you are I/O-bound but not so much if you are CPU
bound, and especially if the whole database is memory resident. I'd
try something like random_page_cost = seq_page_cost = 0.1 for
starters, or whatever values were working for you in 8.1, but the
sweet spot may be higher or lower.
...Robert
From | Date | Subject | |
---|---|---|---|
Next Message | Suvankar Roy | 2009-08-03 04:57:58 | Re: Greenplum MapReduce |
Previous Message | Chris | 2009-08-03 02:04:58 | Re: Greenplum MapReduce |