Re: [HACKERS] EXPLAIN ANALYZE on 8.2 - Mailing list pgsql-performance

From Florian Weimer
Subject Re: [HACKERS] EXPLAIN ANALYZE on 8.2
Date
Msg-id 82irgdnxuz.fsf@mid.bfk.de
Whole thread Raw
In response to Re: [HACKERS] EXPLAIN ANALYZE on 8.2  ("Simon Riggs" <simon@2ndquadrant.com>)
List pgsql-performance
* Simon Riggs:

>> I think the best option is setitimer(), but it's not POSIX so
>> platform support is going to be patchy.
>
> Don't understand that. I thought that was to do with alarms and
> signals.

You could use it for sampling.  Every few milliseconds, you record
which code is executing (possibly using a global variable which is set
and reset accordingly).  But this might lead to tons of interrupted
system calls, and not all systems mask them, so this might not be an
option for the PostgreSQL code base.

On the other hand, if performance in more recent Linux releases (from
kernel.org) are acceptable, you should assume that the problem will
eventually fix itself.  FWIW, I see the 9x overhead on something that
is close to 2.6.17 (on AMD64/Opteron), so this could be wishful
thinking. 8-(

--
Florian Weimer                <fweimer@bfk.de>
BFK edv-consulting GmbH       http://www.bfk.de/
Kriegsstraße 100              tel: +49-721-96201-1
D-76133 Karlsruhe             fax: +49-721-96201-99

pgsql-performance by date:

Previous
From: Martijn van Oosterhout
Date:
Subject: Re: [HACKERS] EXPLAIN ANALYZE on 8.2
Next
From: "Merlin Moncure"
Date:
Subject: Re: New to PostgreSQL, performance considerations