Re: Memory Sizing Advice

From: Pat <pat.casey_at_service-now.com>
Date: Thu, 8 May 2008 20:46:14 -0700 (PDT)
Message-ID: <12a7f1d9-6dce-4ba5-9d41-73c18ab0dd7e@y21g2000hsf.googlegroups.com>


On May 8, 8:31 pm, "Ana C. Dent" <anaced..._at_hotmail.com> wrote:
> Pat <pat.ca..._at_service-now.com> wrote in news:c3b1b601-abd9-4d2c-acaa-
> e4d947b44..._at_d77g2000hsb.googlegroups.com:
>
> > Working set is
> > hard to estimate, but in analagous 100G systems I see a buffer hit
> > ratio or around 70% with a 2.3 G SGA (about as big as I can get it on
> > a 32 bit OS). Based on this and a fair amount of spitball analysis I'd
> > guess the working set on the new box to be around 8-10G.
>
> Choose any hit ratio
>
> http://www.oracledba.co.uk/tips/choose.htm
>
> Using BCHR for anything other than pure wishful dreaming,
> is like using Tarot cards to select your spouse.
>
> There is NO, NO, NO relationship between raw database size (200GB)
> and database performance as measured by ANY metric!
>
> Good Luck on your search for the Holy Grail!

What I want to do is, as much as possible, serve data out of cache. On analogous 32 bit boxes with a 2.3G SGA, I'm seeing an awful lot of physical IOs, enough that many of my queries are spending > 50% of their time in IO wait.

The classic solution to this is:

add more memory

What i want to know is if there's a potential downside to throwing memory at the problem. I have the hardware budget to buy an aweful lot of memory, but I don't want to spend it there if there if it'll be counterproductive. Received on Thu May 08 2008 - 22:46:14 CDT

Original text of this message