Re: testing memory parameters
Date: Mon, 17 Mar 2008 08:36:44 -0700
Serge Rielau wrote:
>> Below what value is BCHR "bad"?
> In DB2 I treat any OLTP app with less than 90% hit ratio as troubled.
> Anything less than 80% I'd call a mess requiring some serious explanation.
Because, in DB2, you don't have sufficient instrumentation to look inside and see what is happening.
>> Above what value is BCHR "good"?
> There is no good (we agree here). There only is "not obviously bad".
> Just like a car whose engine check light is off isn't by definition a
> good car.
There is only one measure that matters: Users complaining that things are taking too much time. Those users might be on a website, they might be running reports, they might be a backup or batch process. But that is the only true measure and ratios will not help you in any way.
>> What are the exceptions to the exceptions?
> Too many $ spent on memory.
No matter how much I spend I still can't remember anything.
> You never can have clear rules. You only can have indicators. The
> ability to separate an indicator from a rule is part of teh job.
> I don't think we are on opposing ends here. I just find the word
> "useless" swinging the pendulum way too far to the other end.
While I agree with your statement, and while it may be true that in DB2 ratios are what you have to work with, take a good look at ASH, AWR, and the wait interface in Oracle and you will see that we have moved a long way away from the Oracle 6 and 7 days when tuning by ratios and tuning by random number generation had value (if they ever did).
I cannot think of a single highly respected tuning expert in the Oracle community ... not Jonathan Lewis ... not Cary Millsap ... not Jeremiah Wilton ... not Graham Wood ... who would suggest BHCR as an important metric.
-- Daniel A. Morgan Oracle Ace Director & Instructor University of Washington damorgan_at_x.washington.edu (replace x with u to respond) Puget Sound Oracle Users Group www.psoug.orgReceived on Mon Mar 17 2008 - 10:36:44 CDT