Re: 8.3 / 8.2.6 restore comparison - Mailing list pgsql-hackers

From Joshua D. Drake
Subject Re: 8.3 / 8.2.6 restore comparison
Date
Msg-id 20080222092606.0debcf59@commandprompt.com
Whole thread Raw
In response to Re: 8.3 / 8.2.6 restore comparison  (Luke Lonergan <llonergan@greenplum.com>)
Responses Re: 8.3 / 8.2.6 restore comparison
List pgsql-hackers
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Tue, 19 Feb 2008 18:54:38 -0800
Luke Lonergan <llonergan@greenplum.com> wrote:

> Bah.
> 
> It's the stuff in the format translation path and conversion to/from
> datums that is the bottleneck.

So how do we fix that? This performance is pretty anemic.

> 
> We sped up COPY TO recently by a factor of 10 using similar
> approaches to what we did for COPY FROM in the past.  There's a
> format conversion that is the culprit.

Any chance that code (in whatever shape) could be released for review
of applicability to core?

> 
> We routinely get about 12 MB/s of heap insertion rate per CPU core
> and it's CPU bound.

That's embarrassing on today's hardware don't you think? Or am I missing
something entirely?

> 
> You can peek in on what's happening using gstack on Linux, or the gdb
> "attach and print stacktrace" approach for a crude profile.
> 

Yeah, that... is beyond my abilities. Well reading it is anyway. I can
provide any information people want though. Tom? Greg? Andrew? Somebody?
What information do you want from me to help you track this down?

Sincerely,

Joshua D. Drake

- -- 
The PostgreSQL Company since 1997: http://www.commandprompt.com/ 
PostgreSQL Community Conference: http://www.postgresqlconference.org/
Donate to the PostgreSQL Project: http://www.postgresql.org/about/donate
PostgreSQL SPI Liaison | SPI Director |  PostgreSQL political pundit

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)

iD8DBQFHvwWuATb/zqfZUUQRAucvAKCAXMlS25oLrTD5bbdIuy6/qq61OgCgn4NG
4I2V5WgSMgr24TAM3CsIO50=
=AaB/
-----END PGP SIGNATURE-----

pgsql-hackers by date:

Previous
From: Robert Lor
Date:
Subject: Proposed changes to DTrace probe implementation
Next
From: Tom Lane
Date:
Subject: Re: Full-text search default vs specified configuration