Re: Transaction Snapshot Cloning

On Fri, 2008-03-28 at 11:33 -0400, Chris Browne wrote:
> bruce(at)momjian(dot)us (Bruce Momjian) writes:
> > Added to TODO:
> >
> > * Allow one transaction to see tuples using the snapshot of another
> > transaction
> >
> > This would assist multiple backends in working together.
> > http://archives.postgresql.org/pgsql-hackers/2008-01/msg00400.php
>
> FYI, code for this is presently available on pgFoundry, at:
> <http://pgfoundry.org/projects/snapclone/>.
>
> I have some benchmarking scripts to commit to it, once I get added to
> the project. (Simon? :-))
>
> FYI, preliminary testing, on a machine hooked up to an EMC CX700 disk
> array is showing that, if I do concurrent dumps of 4 tables, it gives
> a bit better than a 2x speedup over dumping the four tables serially,
> so there's definitely some fruit here.

Oh, very cool. Thanks for testing.

I'll add you now - must have missed that, sorry.

My view is that the TODO item is still needed because we want to work
this into the backend more fully.

--
Simon Riggs
2ndQuadrant http://www.2ndQuadrant.com

PostgreSQL UK 2008 Conference: http://www.postgresql.org.uk