From: | Christian Ramseyer <rc(at)networkz(dot)ch> |
---|---|
To: | Jason Tan Boon Teck <tanboonteck(at)gmail(dot)com> |
Cc: | pgsql-novice(at)postgresql(dot)org |
Subject: | Re: Running on multiple core server |
Date: | 2010-10-14 00:59:34 |
Message-ID: | 4CB655F6.7080501@networkz.ch |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-novice |
On 10/13/2010 06:03 AM, Jason Tan Boon Teck wrote:
>
> Why doesn't the OS make use of the other 3 cores - assigning one to
> Apache and another to PostgreSQL when only 1 client is querying?
>
The scheduler in your Linux kernel decided that it was better to run
both these tasks on the same core. This may have advantages regarding
CPU register, cache or pipeline usage.
Usually, CPU processing power isn't the bottleneck in database
applications, I/O (i.e. how fast can be read from/written to storage)
tends to be the limiting factor. So it's quite likely that running on
two cores would either not improve the total runtime or even slow your
application down.
> Is there anything I can do to tweak?
>
Look for an utility named taskset, usually contained in the schedutils
package. You can use it to set the CPU affinity of processes to specific
CPUs, e.g.
taskset -c 3,4 -p a1,a2,a3 #--> your Apaches with pid a1-3 are now bound
to CPUs 3 & 4.
You'll need to RTFM a bit though to find out how/if it can deal with the
child processes Postgres spawns, I've never tried to do that (or
anything with this utility at all for that matter).
Hope that helps
Christian
From | Date | Subject | |
---|---|---|---|
Next Message | Bruce Momjian | 2010-10-14 16:24:49 | Re: Not quite there yet |
Previous Message | Mladen Gogala | 2010-10-13 12:20:59 | Re: Pgstatspack errors in log files |