Thread: Current wisdom wrt fsm on 8.0
What is the current wisdom on setting the fsm variables
for 8.0? How is it different from 7.4? Or is it?
I am assuming these are the values that changed with
Jan's changes. If not what were those variables?
(I can't seem to track down the threads on these things.)
--elein
============================================================
elein@varlena.com Varlena, LLC www.varlena.com
PostgreSQL Consulting, Support & Training
PostgreSQL General Bits http://www.varlena.com/GeneralBits/
=============================================================
I have always depended on the [QA] of strangers.
elein <elein@varlena.com> writes:
> What is the current wisdom on setting the fsm variables
> for 8.0? How is it different from 7.4? Or is it?
Same as before.
> I am assuming these are the values that changed with
> Jan's changes. If not what were those variables?
No, I can't think of any 8.0 changes that would influence how you set
FSM size. There's been some speculation that with the new ARC code,
it might be worth increasing shared_buffers to larger values than were
useful before. I haven't seen any experiments to back that up though.
There are also some brand-new GUC variables that you can twiddle to
influence the background writer's behavior; and the existence of the
bgwriter might affect your thinking about how often checkpoints need
to occur.
pgsql-performance would probably be the most appropriate venue if you
want to discuss this more.
regards, tom lane
tgl@sss.pgh.pa.us (Tom Lane) writes: > elein <elein@varlena.com> writes: >> What is the current wisdom on setting the fsm variables >> for 8.0? How is it different from 7.4? Or is it? > > Same as before. > >> I am assuming these are the values that changed with Jan's changes. >> If not what were those variables? > > No, I can't think of any 8.0 changes that would influence how you > set FSM size. There's been some speculation that with the new ARC > code, it might be worth increasing shared_buffers to larger values > than were useful before. I haven't seen any experiments to back > that up though. It's also likely that you should expect to get better effectiveness even out of smaller amounts of shared buffers, because they won't get "trashed" by vacuums the way they used to be. Well, overall, the removal of ways of "trashing" cache should make it more effective at all sorts of sizes. But I digress... -- let name="cbbrowne" and tld="cbbrowne.com" in String.concat "@" [name;tld];; http://www.ntlug.org/~cbbrowne/linuxxian.html A VAX is virtually a computer, but not quite.