Re: pg9.0.3 explain analyze running very slow compared to a different box with much less configuration

From: Marti Raudsepp <marti(at)juffo(dot)org>
To: DM <dm(dot)aeqa(at)gmail(dot)com>
Cc: Achilleas Mantzios <achill(at)matrix(dot)gatewaynet(dot)com>, pgsql-performance(at)postgresql(dot)org
Subject: Re: pg9.0.3 explain analyze running very slow compared to a different box with much less configuration
Date: 2011-03-25 08:30:00
Message-ID: AANLkTinoYN6vDJ-CSgUqi+dJaLrpg_72AyfNoc1FO4cV@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

2011/3/25 DM <dm(dot)aeqa(at)gmail(dot)com>:
> gettimeofday() on my new box is slow, after further research we found that,
> when we set ACPI=Off, we got a good clock performance even the explain
> analyze gave approximately gave the right values, but the hyperthreading is
> off.

Disabling ACPI also disables most CPU power management, so that
explains why you get a stable TSC that way. But that's not a real fix.

> could you guide me how to set, the parameter current_clocksource to TSC,

You can't "set" it, the kernel will automatically choose TSC, if it's
stable, at boot time; see messages in dmesg.

A better way to disable power management on CentOS is to disable the
'cpuspeed' service.

Note that this is not necessary for newer CPUs; Intel Nehalem and AMD
Phenom series have a stable TSC even with power management enabled.

Regards,
Marti

In response to

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Adarsh Sharma 2011-03-25 08:55:29 Re: Why Index is not used
Previous Message Chetan Suttraway 2011-03-25 08:14:27 Re: Why Index is not used