Re: Enterprise slow, Standard fast?

From: DA Morgan <damorgan_at_x.washington.edu>
Date: Mon, 08 Nov 2004 00:53:02 -0800
Message-ID: <1099903920.696318@yasure>

Arska wrote:

> Mladen Gogala wrote:
>

>> On Sun, 07 Nov 2004 20:38:32 +0200, Arska wrote:
>>
>>
>>> Now we need to tune every job again, parallel execution in mind. Or deny
>>> parallel execution for these jobs. (How?)
>>
>>
>>
>> No! You should just tune the buffer cache hit ratio. When it is close to
>> 100%, your performance is good. Make sure that BCHR reports are compiled
>> and run every morning. The job of a DBA is to maintain the highest cache
>> hit ratio possible. DBAs are the priests of the buffer cache god. You
>> shouldn't be tuning applications, but the instance. Users are not running
>> applications, they're running instance, so that's what you tune. That
>> means that all you have to do is to fiddle and twist the buffer cache
>> parameters and everything will be good. Make sure that 99% of cache hit
>> ratio is entered into the official company documentation as the ultimate
>> performance metrics.
>>

>
> Buffer cache is currently 1008 MB, and cache hit ratio is 98.9% after
> running few heavy update jobs. Are you saying I need to buy more RAM and
> increase buffer cache? I dont think so.
> We only need to add "ALTER SESSION DISABLE PARALLEL QUERY" to all jobs,
> which was slow with EE, and magically they are fast again.
>
> Parallel query does add some overhead. Usually it is still faster, but
> not always. In this case it is better not to use it.

If he is I wouldn't take his advice.

You might want to get a copy of:
Why a 99%+ Database Buffer Cache Hit Ratio is Not Ok by Cary Millsap

-- 
Daniel A. Morgan
University of Washington
damorgan_at_x.washington.edu
(replace 'x' with 'u' to respond)

Received on Mon Nov 08 2004 - 02:53:02 CST