| From: | Greg Stark <gsstark(at)mit(dot)edu> |
|---|---|
| To: | John Meinel <john(at)johnmeinel(dot)com> |
| Cc: | Francisco Reyes <lists(at)natserv(dot)com>, pgsql-performance(at)postgresql(dot)org |
| Subject: | Re: why my query is not using index?? |
| Date: | 2004-10-12 01:13:27 |
| Message-ID: | 87fz4k3fig.fsf@stark.xeocode.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-performance |
John Meinel <john(at)johnmeinel(dot)com> writes:
> As Janning mentioned, sometimes sequential scans *are* faster. If the number of
> entries that will be found is large compared to the number of total entries (I
> don't know the percentages, but probably >30-40%),
Actually 30%-40% is unrealistic. The traditional rule of thumb for the
break-even point was 10%. In POstgres the actual percentage varies based on
how wide the records are and how correlated the location of the records is
with the index. Usually it's between 5%-10% but it can be even lower than that
sometimes.
--
greg
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Christopher Browne | 2004-10-12 01:34:44 | Re: IBM P-series machines (was: Excessive context |
| Previous Message | Tom Lane | 2004-10-12 00:28:28 | Re: EXPLAIN ANALYZE much slower than running query normally |