From: Vivek Khera <khera@kcilink.com>
To: pgsql-performance@postgresql.org
Subject: restore time: sort_mem vs. checkpoing_segments
Date: Mon, 15 Sep 2003 15:15:46 -0400
Message-ID: <16230.4066.565194.387700@yertle.int.kciLink.com> (raw)
And the winner is... checkpoint_segments.
Restore of a significanly big database (~19.8GB restored) shows nearly
no time difference depending on sort_mem when checkpoint_segments is
large. There are quite a number of tables and indexes. The restore
was done from a pg_dump -Fc dump of one database.
All tests with 16KB page size, 30k shared buffers, sort_mem=8192, PG
7.4b2 on FreeBSD 4.8.
3 checkpoint_segments restore time: 14983 seconds
50 checkpoint_segments restore time: 11537 seconds
50 checkpoint_segments, sort_mem 131702 restore time: 11262 seconds
There's an initdb between each test.
For reference, the restore with 8k page size, 60k buffers, 8192
sort_mem and 3 checkpoint buffers was 14777 seconds.
It seems for restore that a larger number of checkpoint buffers is the
key, especially when dealing with large numbers of rows in a table.
I notice during the restore that the disk throughput triples during
the checkpoint.
The postgres data partition is on a 14-spindle hardware RAID5 on U320
SCSI disks.
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients from your mail client; the link fills in
the headers below: reply via email
Some clients, webmail especially, drop In-Reply-To and References from
a link, and the reply then starts a new thread: use the mbox below.
To: pgsql-performance@postgresql.org
Cc: khera@kcilink.com
Subject: Re: restore time: sort_mem vs. checkpoing_segments
In-Reply-To: <16230.4066.565194.387700@yertle.int.kciLink.com>
References: <16230.4066.565194.387700@yertle.int.kciLink.com>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox