Re: insert performance riddle - Mailing list pgsql-general

From Ed L.
Subject Re: insert performance riddle
Date
Msg-id 200508111529.29046.pgsql@bluepolka.net
Whole thread Raw
In response to Re: insert performance riddle  (Michael Fuhr <mike@fuhr.org>)
Responses Re: insert performance riddle  (Steve Wormley <steve@wormley.com>)
Re: insert performance riddle  (Michael Fuhr <mike@fuhr.org>)
Re: insert performance riddle  ("Greg Sabino Mullane" <greg@turnstep.com>)
List pgsql-general
On Thursday August 11 2005 1:37 pm, Michael Fuhr wrote:
> Have you done any client-side tests that eliminate Perl?  I'd
> suggest writing a little C program so you can measure libpq's
> performance without the extra layers of Perl and DBI/DBD::Pg.
>  Test both local (Unix socket) and network (IPv4 or IPv6
> socket) connections.

Michael, you nailed it again.  My libpq test C program delivered
between 2400 QPS and 5000 QPS vs ~10 QPS for DBI/DBD::Pg on this
box.

It remains unclear to me why the same DBI/DBD::Pg client code
would deliver performance 2-3 orders of magnitude better on
other roughly comparable or inferior boxes.

What would be the recommended replacement for DBI/DBD::Pg as a
perl interface?

Thanks,
Ed


pgsql-general by date:

Previous
From: Alvaro Herrera
Date:
Subject: Re: plphp crashing server
Next
From: Tom Lane
Date:
Subject: Re: No PUBLIC access by default?