Re: Posix AIO in new Red Hat Linux - Mailing list pgsql-hackers

From nconway@klamath.dyndns.org (Neil Conway)
Subject Re: Posix AIO in new Red Hat Linux
Date
Msg-id 20020330183648.GC25612@klamath.dyndns.org
Whole thread Raw
In response to Re: Posix AIO in new Red Hat Linux  (Greg Copeland <greg@CopelandConsulting.Net>)
Responses Re: Posix AIO in new Red Hat Linux  (Greg Copeland <greg@CopelandConsulting.Net>)
List pgsql-hackers
On Sat, Mar 30, 2002 at 09:59:11AM -0600, Greg Copeland wrote:
> It doesn't really say, however, it makes me wonder if it's SGI's KAIO
> (http://oss.sgi.com/projects/kaio/) effort which is reported to provide
> up to 35% performance improvement for heavily I/O bound applications.

I don't think it is. I think the RH announcement refers to their intent
to quickly stabilize some of the upcoming Linux 2.5 code and package it
separately for "enterprise" customers. If that's correct, the Linux 2.5
asynchronous I/O implementation will likely be Ben LaHaise's:
http://www.kvack.org/~blah/aio/

I agree with Greg -- it looks like it could definately improve Pg's
performance.

Cheers,

Neil

-- 
Neil Conway <neilconway@rogers.com>
PGP Key ID: DB3C29FC


pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: PGSTAT start failure probably shouldn't disable Postgres
Next
From: Jessica Perry Hekman
Date:
Subject: Re: timeout implementation issues