Re: Dealing with big tables - Mailing list pgsql-performance

From Matthew
Subject Re: Dealing with big tables
Date
Msg-id Pine.LNX.4.58.0712031032220.3731@aragorn.flymine.org
Whole thread Raw
In response to Re: Dealing with big tables  (Mark Kirkwood <markir@paradise.net.nz>)
List pgsql-performance
On Mon, 3 Dec 2007, Mark Kirkwood wrote:

> > And I cannot use some index organized table or table partitioned by
> > From :) because there are at least 2 similar indexes by which queries
> > can be executed - From and To.

> This makes things a bit tough. One trick is to vertically partition the
> table into two new tables - with "From" in one and "To" in the other...
> then you can (horizontally) partition or cluster on each of these
> columns separately.

Or you could even commit a capital sin and have several copies of the same
table, sorted by different columns. Just remember to select from the
correct table to get the performance, and to write all changes to all the
tables! Kind of messes up transactions and locking a little though.

Matthew

--
No, C++ isn't equal to D. 'C' is undeclared, so we assume it's an int,
with a default value of zero.  Hence, C++ should really be called 1.
-- met24, commenting on the quote "C++ -- shouldn't it be called D?"

pgsql-performance by date:

Previous
From: Tom Lane
Date:
Subject: Re: EXPLAIN ANALYZE time calculations
Next
From: Paul Lindner
Date:
Subject: Re: Training Recommendations