From: | Alvaro Herrera <alvherre(at)commandprompt(dot)com> |
---|---|
To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
Cc: | Greg Stark <gsstark(at)mit(dot)edu>, Pierre C <lists(at)peufeu(dot)com>, Greg Smith <greg(at)2ndquadrant(dot)com>, Dave Crooke <dcrooke(at)gmail(dot)com>, Paul McGarry <paul(at)paulmcgarry(dot)com>, pgsql-performance(at)postgresql(dot)org |
Subject: | Re: shared_buffers advice |
Date: | 2010-03-16 22:20:58 |
Message-ID: | 20100316222058.GI3037@alvh.no-ip.org |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
Tom Lane escribió:
> Alvaro Herrera <alvherre(at)commandprompt(dot)com> writes:
> > Tom Lane escribi:
> >> Reorder to what, though? You still have the problem that we don't know
> >> much about the physical layout on-disk.
>
> > Well, to block numbers as a first step.
>
> fsync is a file-based operation, and we know exactly zip about the
> relative positions of different files on the disk.
Doh, right, I was thinking in the sync-file-range kind of API.
> > We had a customer that had a
> > performance problem because they were inserting lots of data to TOAST
> > tables, causing very frequent extensions. I kept wondering whether an
> > allocation policy that allocated several new blocks at a time could be
> > useful (but I didn't try it). This would also alleviate fragmentation,
> > thus helping the physical layout be more similar to logical block
> > numbers.
>
> That's not going to do anything towards reducing the actual I/O volume.
> Although I suppose it might be useful if it just cuts the number of
> seeks.
Oh, they had no problems with I/O volume. It was relation extension
lock that was heavily contended for them.
--
Alvaro Herrera http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support
From | Date | Subject | |
---|---|---|---|
Next Message | Tom Lane | 2010-03-16 22:25:26 | Re: shared_buffers advice |
Previous Message | Tom Lane | 2010-03-16 21:58:34 | Re: shared_buffers advice |