[nycbug-talk] Memory sizing

[Pete Wright]

On Sat, Apr 22, 2006 at 07:26:25PM -0400, Marc Spitzer wrote:
> On 4/22/06, Pete Wright <pete at nomadlogic.org> wrote:
> > as an aside, i have another postgres DB that we are running out of a
> > memory FS.  it's execellent, although took a bit of work on our end to
> > architect a solution that would work this way.  the database in the memfs
> > partition does not persist between reboot's - it populates itself off
> > an oracle server (and through some other methods) between reboots.  it
> > has been very successfull performance wise, although granted such a
> > solution may not fit everyone's environment.
> 
> what kind of speedup did you see vs a normal(disk based) database?
> 
> marc

i don't have the numbers handy but it was pretty huge in our case.  this
is a relativy small DB (~512M) that does a lot of small read/writes from
a large pool of workstations and nodes in our render farm.  because the
data did not need to persist between reboots it worked out well in our
case.  

-p

-- 
~~oO00Oo~~
Peter Wright
pete at nomadlogic.org
www.nomadlogic.org/~pete
310.869.9459