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

Re: Writing Commit Status hint bits (was Re: [HACKERS] Constant WAL replay)

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Simon Riggs <simon(at)2ndquadrant(dot)com>
Cc: pgsql-patches(at)postgresql(dot)org, pgsql-hackers(at)postgresql(dot)org
Subject: Re: Writing Commit Status hint bits (was Re: [HACKERS] Constant WAL replay)
Date: 2005-07-22 13:40:09
Message-ID: 28960.1122039609@sss.pgh.pa.us (view raw or flat)
Thread:
Lists: pgsql-hackerspgsql-patches
Simon Riggs <simon(at)2ndquadrant(dot)com> writes:
> In general, the hint bits are good. In *some* cases, not. I still seek
> control over that as a designer.

> Specifically, the scenario I want to optimize is this:
> - we load a table with lots of real time measurement data, as one child
> out of a large number of similar child tables
> - we then immediately create summary tables from the measurements
> - after this the detailed data is only sporadically accessed, if ever
> - detail data is dropped from the database after a few weeks
> - the majority of the database is detail data, so those tables are never
> vacuumed since no rows are ever deleted from those tables (the tables
> are dropped) nor is access sufficiently frequent to make it sensible to
> set hint bits - hence *no* complete database vacuum is run on a regular
> basis

> In this design, the second step causes the whole detailed data table to
> be written out to the database (again) immediately after loading. I
> would like to be able to prevent that.

This application seems sufficiently off the beaten path to not be worth
solving with a knob as klugy and potentially dangerous as
suppress_hint_bits.

A better way to avoid the write pass, if you must, is to do the summary
table creation in the same transaction that loads the data.

			regards, tom lane

In response to

Responses

pgsql-hackers by date

Next:From: Tom LaneDate: 2005-07-22 13:53:24
Subject: Re: [PATCHES] Roles - SET ROLE Updated
Previous:From: Stephen FrostDate: 2005-07-22 13:17:06
Subject: Re: [PATCHES] Roles - SET ROLE Updated

pgsql-patches by date

Next:From: Tom LaneDate: 2005-07-22 13:53:24
Subject: Re: [PATCHES] Roles - SET ROLE Updated
Previous:From: Stephen FrostDate: 2005-07-22 13:17:06
Subject: Re: [PATCHES] Roles - SET ROLE Updated

Privacy Policy | About PostgreSQL
Copyright © 1996-2014 The PostgreSQL Global Development Group