Re: Indexes?

From: Stephan Szabo <sszabo(at)megazone(dot)bigpanda(dot)com>
To: Mike Leahy <mgleahy(at)fes(dot)uwaterloo(dot)ca>
Cc: pgsql-general(at)postgresql(dot)org
Subject: Re: Indexes?
Date: 2003-10-15 04:10:55
Message-ID: 20031014204201.T89503@megazone.bigpanda.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

On Tue, 14 Oct 2003, Mike Leahy wrote:

> CREATE INDEX tbl_censo_poblacion_1993_manzana_idx ON tbl_censo_poblacion_1993
> ( dubicacion, zona, manzana );
>
> Then I try a query such as:
>
> select count(*) as POBLACION, (dubicacion || zona || manzana) as COD_MANZANA
> from tbl_censo_poblacion_1993 group by dubicacion, zona, manzana;
>
> The results from explain indicate that a sequential scan is used (as far as I
> can tell). I tried adding where statement:
>
> select count(*) as POBLACION, (dubicacion || zona || manzana) as COD_MANZANA
> from tbl_censo_poblacion_1993 where dubicacion <> '' and zona <> '' and
> manzana <> '' group by dubicacion, zona, manzana;
>
> The EXPLAIN analysis results appear as follows:
>
> Aggregate (cost=847587.90..879024.28 rows=251491 width=27) (actual
> time=272782.00..279458.00 rows=21459 loops=1)
> -> Group (cost=847587.90..872737.01 rows=2514911 width=27) (actual
> time=272782.00..278546.00 rows=2553015 loops=1)
> -> Sort (cost=847587.90..853875.18 rows=2514911 width=27) (actual
> time=272782.00..274533.00 rows=2553015 loop
> s=1)
> Sort Key: dubicacion, zona, manzana
> -> Seq Scan on tbl_censo_poblacion_1993 (cost=0.00..328346.76
> rows=2514911 width=27) (actual time=0.00. .189570.00 rows=2553015 loops=1)

Note how many rows it's estimating (and actually) getting matches for.
It appears to be matching nearly every row so indexes aren't going to
really help much here for PostgreSQL since it still has to go back to the
actual table to find out whether or not the row is visible to your
transaction.

What does vacuum verbose tbl_censo_poblacion_1993; say? Specifically, how
many pages does the table have?

> As an aside, is there any way to increase the amount of memory allocated to
> the postmaster.exe process? It seems to me that if I could add more than 4MB
> that it has by default, then maybe that could increase the performance.

You might want to raise both shared_buffers and sort_mem from their
default values in postgresql.conf.

For more information you should probably read:
http://www.varlena.com/varlena/GeneralBits/Tidbits/perf.html
http://www.varlena.com/varlena/GeneralBits/Tidbits/annotated_conf_e.html

In response to

  • Indexes? at 2003-10-15 01:57:05 from Mike Leahy

Responses

Browse pgsql-general by date

  From Date Subject
Next Message David So 2003-10-15 04:34:30 Cygwin? or paid version from SRA?
Previous Message Tom Lane 2003-10-15 04:07:12 Re: PLPERL function error - utf-8 to iso8859-1