Re: how small to split a table?

From: "Heiko W(dot)Rupp" <heiko(dot)rupp(at)redhat(dot)com>
To: Vivek Khera <vivek(at)khera(dot)org>
Cc: "Pgsql-Performance ((((E-mail))))" <pgsql-performance(at)postgresql(dot)org>
Subject: Re: how small to split a table?
Date: 2007-03-20 15:20:42
Message-ID: 6AA70D81-4D28-44F9-A776-A401899517B0@redhat.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance


> My question is how small to make each inherited piece? If I do
> modulo 10, then each sub-table will be between 32 and 34 million
> rows today based on current distribution.

You might try this with various sizes.
I did some testing lateley and found out that insert performance -
even if only inserting into one partition through the master
table abould halfed the speed with 4 partitions and made a 50%
increase for 2 partitions.
Please note: this is not representative in any kind!

So while it might be cool in your case to have e.g. one partition per
month, this might slow inserts down too much, so
that a different number of partitions could be better. The same
applies for queries as well (here perhaps in the other
direction).

--
Heiko W.Rupp
heiko(dot)rupp(at)redhat(dot)com, http://www.dpunkt.de/buch/3-89864-429-4.html

In response to

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Joshua D. Drake 2007-03-20 15:41:50 Re: SATA RAID: Promise vs. 3ware
Previous Message Vivek Khera 2007-03-20 15:02:32 how small to split a table?