On Mon, Jul 07, 2003 at 04:01:12PM -0600, scott.marlowe wrote:
>
> This increase in buffers isn't free, since it will now cost Postgresql the
> overhead of manageing said buffers, plus they're in unix shared memory,
> which isn't all that fast compared to kernel level cache.
Also, I learned through painful experience that it gets slower as it
has to manage large data sets. After some testing, we attempted a 2G
buffer size once, on the grounds that we had 12 G. At first, we saw
some happy results, but after a couple days, things got painfully
slow. I eventually concluded that once the buffer was full, managing
it was making the system crawl.
A
--
----
Andrew Sullivan 204-4141 Yonge Street
Liberty RMS Toronto, Ontario Canada
<andrew@libertyrms.info> M2P 2A8
+1 416 646 3304 x110