Re: Runtime variations during day

Carlos Henrique Reimer <carlos(dot)reimer(at)opendb(dot)com(dot)br> wrote:

> Anyway it does not seam related to statistics as the query plan
> is exactly the same for both scenarios, morning and evening:

> Morning:

>   Index Scan using pagpk_aux_mes, pagpk_aux_mes, pk_cadpag,
>     pk_cadpag, pk_cadpag, pagchavefunc00 on cadpag

> Evening:

>   Seq Scan on cadpag

Those don't look exactly the same to me.

> I've used this query just as an example but the general feeling
> is that everything takes more time to process in the evening.
> Evening is also the period with more tasks in the the database.

Well, if the resources are split among more processes, each process
will take longer to run.  No surprise there.  Of course, there
could be some interaction -- queries which do more work because
statistics are out of date could contribute to the slowdown of
everything else.

--
Kevin Grittner
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company