| From: | Merlin Moncure <mmoncure(at)gmail(dot)com> |
|---|---|
| To: | Humair Mohammed <humairm(at)hotmail(dot)com> |
| Cc: | pgsql-performance(at)postgresql(dot)org |
| Subject: | Re: Query Performance SQL Server vs. Postgresql |
| Date: | 2010-11-23 14:20:05 |
| Message-ID: | AANLkTinNLVJ_ypOAML4UmfZhKCz+eqstSQWaYbq390iJ@mail.gmail.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-performance |
On Mon, Nov 22, 2010 at 7:12 PM, Humair Mohammed <humairm(at)hotmail(dot)com> wrote:
> I did some further analysis and here are the results:
> work_mem;response_time
> 1MB;62 seconds
> 2MB;2 seconds
> 4MB;700 milliseconds
> 8MB;550 milliseconds
> In all cases shared_buffers were set to the default value of 32MB. As you
> can see the 1 to 2 MB jump on the work_mem does wonders. I probably don't
> need this to be any higher than 8 or 16 MB. Thanks to all for help!
> Humair
work_mem directly affects how the query is planned, because certain
types of plans (hash joins ans large sorts) require memory. raising or
lowering shared_buffers OTOH is very subtle and is not something you
tune to improve the execution of a single query...
merlin
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Craig Ringer | 2010-11-24 00:11:28 | Re: Performance under contention |
| Previous Message | Humair Mohammed | 2010-11-23 00:12:30 | Re: Query Performance SQL Server vs. Postgresql |