| From: | "Luke Lonergan" <llonergan(at)greenplum(dot)com> |
|---|---|
| To: | "Tom Lane" <tgl(at)sss(dot)pgh(dot)pa(dot)us>, "Simon Riggs" <simon(at)2ndquadrant(dot)com> |
| Cc: | "Heikki Linnakangas" <hlinnaka(at)iki(dot)fi>, pgsql-patches(at)postgresql(dot)org, "Eng" <eng(at)intranet(dot)greenplum(dot)com>, "System Engineering" <system_engineering(at)intranet(dot)greenplum(dot)com> |
| Subject: | Re: Page at a time index scan |
| Date: | 2006-05-08 19:09:11 |
| Message-ID: | C084E567.22FA6%llonergan@greenplum.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-patches |
Tom,
On 5/8/06 11:46 AM, "Tom Lane" <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> I made a table of 16M rows with an
> index over a random-data integer column. With a thoroughly disordered
> index (built on-the-fly as the random data was inserted), the time to
> VACUUM after deleting a small number of rows was 615 seconds with
> yesterday's code, 31 seconds today. With a perfectly-ordered index
> (identical table, but CREATE INDEX after all the data is in place), the
> times were about 28 and 26 seconds respectively.
Very impressive! This corroborates findings we've had with index
maintenance in the field - thanks for finding/fixing this.
- Luke
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Alvaro Herrera | 2006-05-08 19:46:54 | Re: [WIP] The relminxid addition, try 3 |
| Previous Message | Tom Lane | 2006-05-08 19:08:19 | Re: [WIP] The relminxid addition, try 3 |