Re: Clock sweep not caching enough B-Tree leaf pages?

From: Peter Geoghegan <pg(at)heroku(dot)com>
To: Robert Haas <robertmhaas(at)gmail(dot)com>
Cc: Andres Freund <andres(at)2ndquadrant(dot)com>, Merlin Moncure <mmoncure(at)gmail(dot)com>, Pg Hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Clock sweep not caching enough B-Tree leaf pages?
Date: 2014-04-16 23:22:59
Message-ID: CAM3SWZQstGL+Njvg00Y3gLKTHz+aNZAikoyxN734hcm9PwirCw@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Wed, Apr 16, 2014 at 7:29 AM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> 2. Improving the choice of which buffers we evict, which is what
> Peter's talking about, or at least what I think he's talking about.
>
> Those things are both important, but they're different, and I'm not
> sure that working on one precludes working on the other. There's
> certainly the potential for overlap, but not necessarily.

That's certainly my primary area of interest here, but there may be
some overlap with other areas, or areas were cooperation turns out to
be appropriate. At the risk of stating the very obvious, if the
caching algorithm is making poor decisions about caching, and leaf
pages are continually swapped in and out of shared_buffers for no good
reason, that is likely to constrain the scalability of the buffer
manager. That could be very significant.

My immediate concern here is getting recognition of the importance of
weighing frequency of access in *some* way.

--
Peter Geoghegan

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Bruce Momjian 2014-04-16 23:24:40 Re: [COMMITTERS] pgsql: Include planning time in EXPLAIN ANALYZE output.
Previous Message Tom Lane 2014-04-16 23:19:44 Re: assertion failure 9.3.4