From: | Bruce Momjian <bruce(at)momjian(dot)us> |
---|---|
To: | Robert Haas <robertmhaas(at)gmail(dot)com> |
Cc: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, Hannu Krosing <hannu(at)2ndquadrant(dot)com>, Peter Hussey <peter(at)labkey(dot)com>, pgsql-performance <pgsql-performance(at)postgresql(dot)org> |
Subject: | Re: Questions on query planner, join types, and work_mem |
Date: | 2011-02-01 03:17:57 |
Message-ID: | 201102010317.p113Hvc07473@momjian.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
Robert Haas wrote:
> On Wed, Aug 11, 2010 at 9:42 PM, Bruce Momjian <bruce(at)momjian(dot)us> wrote:
> > This confused me. ?If we are assuing the data is in
> > effective_cache_size, why are we adding sequential/random page cost to
> > the query cost routines?
>
> See the comments for index_pages_fetched(). We basically assume that
> all data starts uncached at the beginning of each query - in fact,
> each plan node. effective_cache_size only measures the chances that
> if we hit the same block again later in the execution of something
> like a nested-loop-with-inner-indexscan, it'll still be in cache.
>
> It's an extremely weak knob, and unless you have tables or indices
> that are larger than RAM, the only mistake you can make is setting it
> too low.
The attached patch documents that there is no assumption that data
remains in the disk cache between queries. I thought this information
might be helpful.
--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ It's impossible for everything to be true. +
Attachment | Content-Type | Size |
---|---|---|
/rtmp/effective.diff | text/x-diff | 1.0 KB |
From | Date | Subject | |
---|---|---|---|
Next Message | Greg Smith | 2011-02-01 11:53:49 | Re: Configuration for a new server. |
Previous Message | Benjamin Krajmalnik | 2011-01-31 23:55:32 | Configuration for a new server. |