Re: AIX Oracle Tuning

From: Anton Dischner <dischner_at_med.uni-muenchen.de>
Date: 1995/06/01
Message-ID: <dischner-0106951532480001_at_gkc12a.klch.med.uni-muenchen.de>#1/1


Hi Mike,

take a look at hints for the cost based optimizer.

You may get performance gains using:

select /*+ index(<table_name> <index_name>) */ * from <table_name>;
select /*+ first_rows */ ... <- returns first couple of data very fast
select /*+ rule */

You can alter your complete session using the "alter session" command

Make shure to use

analyze table <table_name> estimate statistics; analyze index ...

to use the cost based optimizer.

When you say query only you should use pctfree 5 pctused 90 to use as less as possible datablocks.

See your application developers guide: B1 Performance diagnostic tools

Maybe a use_readv=true in the initXXXX.ora is good for your (good for complete table scans normally).

Hope this helps,

Toni

In article <3q52o0$n3r_at_newsbf02.news.aol.com>, mmorley1_at_aol.com (MMorley1) wrote:

> Looking for help tuning Oracle on an RS6000 running AIX. The environment
> is 10 tables each with potentially 5 million rows. The data is query only
> and it is unlikely that each record will be queried more than once in a 24
> hour period. This would seem to reduce the effects of changes in cache (or
> SGA) size. Is there a way to just cache indexes? Each table has a single
> 24 byte key. We are looking for the highest throughput possible for single
> returned record selects. Thanks in advance for any ideas.
>
> Mike Morley
> Equifax Inc.
> Tampa
 

-- 
A. Dischner
Institut fuer Klinische Chemie
Klinikum Grosshadern
Ludwig Maximilians Universitaet Muenchen
Marchioninistr. 15 81366 Muenchen 49-89-70953202
Germany
Received on Thu Jun 01 1995 - 00:00:00 CEST

Original text of this message