[PostgreSQL-Hackers] Multiple Storage per Tablespace, or Volumes

	Tom Lane
	Mar 9, 2007 at 6:43 am

Greg Smith writes:
With the pgbench workload, a substantial percentage of the buffer cache
ends up pinned.

[ raised eyebrow... ] Prove that. AFAIK it's impossible for the
pgbench queries to pin more than about three or four buffers per backend
concurrently.

regards, tom lane

Search Discussions

Discussion Posts

Greg Smith: With the pgbench workload, a substantial percentage of the buffer cache ends up pinned. From staring at the buffer cache using contrib/pg_buffercache, I believe most of that consists of the index blocks for the records being updated in the accounts table. I just posted a new version of the patch I asked for feedback on at the beginning of this thread, the latest one is at http://westnet.com/~gsmith/content/postgresql/new-patch-checkpoint.txt I've been adjusting it to monitor the same data I

Follow ups

Related Discussions

Discussion Navigation

view	thread \| post
posts	‹ prev \| 35 of 50 \| next ›

Discussion Overview

group	pgsql-hackers
categories	postgresql
posted	Feb 19, '07 at 10:54a
active	Mar 28, '07 at 1:33p
posts	50
users	17
website	postgresql.org...
irc	#postgresql

font	variable	Hotkey: f
user style	avatars	Hotkey: a

[PostgreSQL-Hackers] Multiple Storage per Tablespace, or Volumes

Search Discussions

Discussion Posts

Related Discussions

17 users in discussion

Content

People

Support

Translate