Skip site navigation (1) Skip section navigation (2)

PG block sizes

From: Tom Arthurs <tarthurs(at)jobflash(dot)com>
To: pgsql-performance(at)postgresql(dot)org
Subject: PG block sizes
Date: 2005-02-27 22:32:01
Message-ID: (view raw, whole thread or download thread mbox)
Lists: pgsql-performance
Hi, All

I'm trying to tune a software RAID 0 (striped) on a solaris 9, sparc box.  Currently I'm using a raid 1 (mirrored) array on two discs for the data area,
and I put in 4 new drives last night (all are f-cal).  On the new array I have a width of 4, and used the default interleave factor of 32k.  I believe
a smaller interleave factor may get me better read performance (I'm seeing a bulk load performance increase of about 35% but a 7-8x worse read performance
between the two RAID setups.)

Conventional wisdom is  using an interleave factor < = db default block size gives the best read performance.  I would like to try that (though this testing
is burning a lot of daylight, since I'll have to reload the db every time I remake the RAID.)

Question:  what't the best block size to use for postgresql on solaris?  (I'm using 7.4.5)

pgsql-performance by date

Next:From: Markus SchaberDate: 2005-02-28 08:10:36
Subject: Re: Peformance Tuning Opterons/ Hard Disk Layout
Previous:From: Simon RiggsDate: 2005-02-27 20:25:45
Subject: Re: Possible interesting extra information for explain

Privacy Policy | About PostgreSQL
Copyright © 1996-2017 The PostgreSQL Global Development Group