Re: testing memory parameters

From: Serge Rielau <srielau_at_ca.ibm.com>
Date: Sun, 16 Mar 2008 21:44:56 -0400
Message-ID: <6460opF29k6iqU1@mid.individual.net>


DA Morgan wrote:
> Ana C. Dent wrote:

>> "mr..._at_gmail.com" <mrowkin_at_gmail.com> wrote in news:62772aa8-fdbd-42b7-
>> a8a0-da91957e30e9_at_e6g2000prf.googlegroups.com:
>>
>>> I think that I should consider buffer and library cache hit ratios.
>>> Is this a good way?
>>
>> BUFFER CACHE HIT RATIO ia a mythical indicator of performance.
>> It should be ignored because it is meaningless.

>
> Just to reinforce what Ana wrote ... hit ratios are meaningless
> nonsense promoted by people who still think they can apply Oracle
> 7 bad practices to current versions of Oracle.
>
> If you have one of those books that recommends such nonsense do
> something useful with it. I suggest a recycling bin.
Hmmm.. OK I bite. The only thing worse than admitting ignorance being remaining ignorant....

While I agree that a high buffer pool hit ratio is not a guaranty for a well tuned system, in OLTP the lack thereof tends to be a guaranty for a badly tuned system (oftentimes caused by a lack of a good access path such as indexing).
Thus a good hit ratio is generally a necessary, but not sufficient indicator for a well tunes application.
I thought this was a universal rule for the major DBMS.

Is my reasoning wrong as far as Oracle is concerned? If so where?

Cheers
Serge

-- 
Serge Rielau
DB2 Solutions Development
IBM Toronto Lab
Received on Sun Mar 16 2008 - 20:44:56 CDT

Original text of this message