Carlos Henrique Reimer wrote:
> Another example that could help is this seqscan:
>=20
> explain analyze select sittrib8 from iparq.arript where sittrib8=3D33;
>=20
> In the evening:
> Fri Feb 8 14:00:01 BRST 2013
>=20
> QUERY
> PLAN
> -----------------------------------------------------------------------=
---------------------------------------------------
> Seq Scan on arript (cost=3D100000000.00..100469613.21 rows=3D1 width=3D=
2)
> (actual time=3D198047.253..198047.253 rows=3D0 loops=3D1)
> Filter: (sittrib8 =3D 33)
> Total runtime: 198047.303 ms
> (3 rows)
>=20
>=20
> In the morning:
> Fri Feb 8 10:51:01 BRST 2013
> QUERY
> PLAN
> -----------------------------------------------------------------------=
-------------------------------------------------
> Seq Scan on arript (cost=3D100000000.00..100469607.58 rows=3D1 width=3D=
2)
> (actual time=3D11982.597..11982.597 rows=3D0 loops=3D1)
> Filter: (sittrib8 =3D 33)
> Total runtime: 11982.654 ms
> (3 rows)
When I have seen things such as this, it's because the disks are busier
during the slow hours. Since there's not much activity in the morning,
the disks are idle and can scan the pages much more quickly.
--=20
=C1lvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services
