| From: | Jeff Davis <pgsql(at)j-davis(dot)com> |
|---|---|
| To: | sthomas(at)optionshouse(dot)com |
| Cc: | Mark Thornton <mthornton(at)optrak(dot)com>, pgsql-performance(at)postgresql(dot)org |
| Subject: | Re: Performance of CLUSTER |
| Date: | 2012-06-11 23:53:57 |
| Message-ID: | 1339458837.12042.9.camel@sussancws0025 |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-performance |
On Mon, 2012-06-11 at 08:42 -0500, Shaun Thomas wrote:
> On 06/10/2012 03:20 AM, Mark Thornton wrote:
>
> > 4. 3980922 rows, 1167MB, 276s
> > 5. 31843368 rows, 9709MB, ~ 10 hours
>
> Just judging based on the difference between these two, it would appear
> to be from a lot of temp space thrashing. An order of magnitude more
> rows shouldn't take over 100x longer to cluster, even with GIST. What's
> your maintenance_work_mem setting?
GiST can have a large impact depending on all kinds of factors,
including data distribution.
9.2 contains some important improvements in GiST build times for cases
where the index doesn't fit in memory. Mark, can you please try your
experiments on the 9.2beta and tell us whether that helps you?
Regards,
Jeff Davis
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Tatsuo Ishii | 2012-06-12 08:54:25 | Re: pg_dump and thousands of schemas |
| Previous Message | Jeff Janes | 2012-06-11 16:32:52 | Re: pg_dump and thousands of schemas |