Re: Transparent table partitioning in future version of PG? - Mailing list pgsql-performance

From Simon Riggs
Subject Re: Transparent table partitioning in future version of PG?
Date
Msg-id 1241647680.6109.61.camel@ebony.2ndQuadrant
Whole thread Raw
In response to Re: Transparent table partitioning in future version of PG?  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: Transparent table partitioning in future version of PG?  (Alvaro Herrera <alvherre@commandprompt.com>)
Re: Transparent table partitioning in future version of PG?  (Robert Haas <robertmhaas@gmail.com>)
List pgsql-performance
On Wed, 2009-05-06 at 17:55 -0400, Tom Lane wrote:
> Simon Riggs <simon@2ndQuadrant.com> writes:
> > On Fri, 2009-05-01 at 11:27 -0400, Robert Haas wrote:
> >> The problem has been finding someone who has both the time and the
> >> ability to do the work.
>
> > Unfortunately there has been significant debate over which parts of
> > partitioning need to be improved. My own view is that considerable
> > attention needs to be applied to both the executor and planner to
> > improve matters and that syntax improvements are largely irrelevant,
> > though seductive.
>
> My thought about it is that what we really need is an explicit notion
> of partitioned tables built into the system, instead of trying to make
> the planner re-deduce the partitioning behavior from first principles
> every time it builds a plan for such a table.  Such a notion would
> presumably involve some new syntax to allow the partitioning rule to be
> specified at table creation time.  I agree that the syntax details are a
> minor issue, but the set of possible partitioning rules is certainly a
> topic of great interest.

Agreed. Perhaps I should say then that the syntax needs to express the
requirements of the planner/executor behaviour, rather than being the
main aspect of the feature, as some have suggested.

Hopefully, notions of partitioning won't be directly tied to chunking of
data for parallel query access. Most queries access recent data and
hence only a single partition (or stripe), so partitioning and
parallelism and frequently exactly orthogonal.

--
 Simon Riggs           www.2ndQuadrant.com
 PostgreSQL Training, Services and Support


pgsql-performance by date:

Previous
From: Tom Lane
Date:
Subject: Re: GiST index performance
Next
From: Tom Lane
Date:
Subject: Re: Transparent table partitioning in future version of PG?