Re: the number of child tables --table partitioning - Mailing list pgsql-performance

From Merlin Moncure
Subject Re: the number of child tables --table partitioning
Date
Msg-id CAHyXU0zy0A8Wwk5CYpr9-FXjKQhZ7Zq29TqO+cQWRvk-dwgFzw@mail.gmail.com
Whole thread Raw
In response to Re: the number of child tables --table partitioning  (Ondrej Ivanič <ondrej.ivanic@gmail.com>)
Responses Re: the number of child tables --table partitioning
List pgsql-performance
2011/9/29 Ondrej Ivanič <ondrej.ivanic@gmail.com>:
> Hi,
>
> On 30 September 2011 01:08, Kevin Grittner <Kevin.Grittner@wicourts.gov> wrote:
>>> Is there a suggested number of child tables for table
>>> partitioning,
>>
>> Generally, don't go over about 100 partitions per table.
>
> Having 365 partitions per table is fine...

yeah -- the system was certainly designed to support 'dozens to
hundreds', but 'hundreds of thousands' is simply not realistic.  any
measurable benefit gained from partitioning is going to be var
exceeded by the database having to track so many tables.

btw, partitioning for purposes of performance is a dubious strategy
unless you can leverage non-uniform access patterns of the data or do
other tricks that allow simplification of structures (like removing
'company_id' from all tables and indexes because it's implied by the
partition itself).

merlin

pgsql-performance by date:

Previous
From: pasman pasmański
Date:
Subject: Re: Shortcutting too-large offsets?
Next
From: alexandre - aldeia digital
Date:
Subject: Re: the number of child tables --table partitioning