Message-ID: <4CA7EDA5.4080401@postnewspapers.com.au>
Date: Sun, 03 Oct 2010 10:42:45 +0800
From: Craig Ringer <craig@postnewspapers.com.au>
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 6.1; en-GB;
	rv:1.9.1.12) Gecko/20100914 Thunderbird/3.0.8
MIME-Version: 1.0
To: Richard Troy <rtroy@ScienceTools.com>
CC: pgsql-performance@postgresql.org
Subject: Re: How does PG know if data is in memory?
References: <Pine.LNX.4.33.1010021626280.8681-100000@denzel.in>
In-Reply-To: <Pine.LNX.4.33.1010021626280.8681-100000@denzel.in>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

On 3/10/2010 7:39 AM, Richard Troy wrote:

> I can't speak for modern "OpenVMS", but "back in the day", VMS had a very
> effective memory management strategy which, in effect, made it as if all
> memory was a cache for disk. It did this by means of a mechanism by which
> to identify all potentially reachable disk space. When disk was read in,
> an entry would be made mapping the memory to the disk space from which it
> came - and if it was later updated, the mapping entry was marked "dirty."
> Whenever disk access was contemplated, a check was made to see if it was
> already in memory and if so, it'd provide access to the in-memory copy
> instead of doing the read again. (This also permitted, under some
> circumstances, to reduce write activity as well.)

That's how Linux's memory management works, too, at least if I 
understand you correctly. Pretty much every modern OS does it. Pg is 
reliant on the operating system's disk cache, and has some minimal 
knowledge of it (see effective_cache_size) .

I don't know how shared_buffers management works, but certainly at the 
OS cache level that's what already happens.

-- 
Craig Ringer

Tech-related writing at http://soapyfrogs.blogspot.com/