| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
Home -> Community -> Usenet -> c.d.o.server -> Re: Free Memory in relation to SGA
A dictionary cache hit ratio of anything over 85% is actually pretty good
(you'll never get much more than that, because, unlike the library cache if
the 'keep' procedure is used, you can't pre-load the dictionary cache, and
therefore the first queries after startup are inevitably going to induce
cache misses). A poor library cache hit ratio will be the product of poor
(ie, unshareable) SQL, and increasing the shared pool size isn't going to
help that at all: the only thing that can boost that ratio is re-writing
your apps' code so that it makes more use of bind variables. If you had a
poor reloads to pins ratio, then increasing the size might be of some use,
but not if it's the library cache hit ratio.
What's more, think about it: if 400+ Mb of your *existing* shared pool is sitting around free, it means your shared pool size is already too big. Free memory in a cache is wasted memory.
I'd be thinking about reducing my shared pool and investing the memory elsewhere -the buffer cache, say. Assuming that the SGA as a whole is not already more than about 1/3rd of available RAM.
Regards
HJR
-- ---------------------------------------------- Resources for Oracle: http://www.hjrdba.com =============================== "twerty" <2desp_at_iname.com> wrote in message news:a7714b$j4hfi$1_at_ID-51406.news.dfncis.de...Received on Tue Mar 19 2002 - 08:49:51 CST
> Hi,
>
> When I do a select * from v$sgastat, I find that my free memory in the
> SGA amounts to 484MB. As my library cache ratio hit and dictionary cache
> ratio hit are not as high as I would like them to be (between 90-94%), I
am
> planning to use this particular free memory in the SGA (484 MB). My
question
> is, does Oracle actually use this free memory in the SGA for its work
> processes? Or can I safely say that it is not a problem assigning the
> remaining say, 400 or 484MB of free space to my Shared Pool Memory and the
> Buffer Cache (which isn't of a very high quality either).
>
> Many Thanks in Advance for the answers.
>
>
![]() |
![]() |