Re: Table Partitions / Partial Indexes

From: Mike C <smith(dot)not(dot)western(at)gmail(dot)com>
To: Simon Riggs <simon(at)2ndquadrant(dot)com>
Cc: pgsql-performance(at)postgresql(dot)org
Subject: Re: Table Partitions / Partial Indexes
Date: 2005-12-13 23:54:08
Message-ID: bd0eabd0512131554v451a99f2u4ac64e403e9b6b49@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

On 12/14/05, Simon Riggs <simon(at)2ndquadrant(dot)com> wrote:
>
> Maybe not for queries, but if you use a date range then you never need
> to run a DELETE and never need to VACUUM.
>
> You could split the data into two-day chunks.

That's an interesting idea, thanks.

> Am I using a horrid method for partitioning the data? (% 10)
>
> No, but what benefit do you think it provides. I'm not sure I see...

I was trying to get both the indexes to be smaller without loosing
selectivity, and make any table scans/index scans faster from having to read
less data.

> Should there be that big of an improvement for multiple tables given
> > that all the data is still stored on the same filesystem?
>
> You could store partitions in separate tablespaces/filesystems.
>

Ideally that's what I would do, but to make the most of that I would have to
have a dedicated RAID setup for each partition right? (Which is a bit pricey
for the budget).

Cheers,

Mike

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message Luke Lonergan 2005-12-14 01:54:38 Re: Should Oracle outperform PostgreSQL on a complex
Previous Message Simon Riggs 2005-12-13 23:28:49 Re: Should Oracle outperform PostgreSQL on a complex