Re: Calling out DBA tuning experts

From: Connor McDonald <mcdonald.connor.cs_at_bhp.com.au>
Date: Mon, 11 Jan 1999 11:34:18 +0800
Message-ID: <3699713A.58A4@bhp.com.au>

Brian + Susan McCracken wrote:
>
> I have recently upgraded from 7.2.3 to 7.3.4
>
> The batch is now taking 50% longer to complete.
>
> The tkprof out put from both 7.3.4 and 7.2.3 show almost equivalent times
> for total elapsed time despite the fact that 7.3.4 takes 50% longer.
>
> This mark up is consistent throughout.
>
> I believe the proble may be related to a relationship beteen the Oracle
> processes and the HP-UX 10.20 OS.
>
> A bstat/estat was produced for both versions of the database and thye
> significant difference was detected in the session event
>
> rdbms ipc message statistics which was 6 times greater in overall elapsed
> time
>
> Does anyone have any info on this statistic and what it means.

(Assuming CBO) I have found that 7.3 tended to invoke the hash join mechanism where I didn't really want it to...

Look thru your trace and explain the big running SQL's - if they come back with hash join, then your options then become:

Recode the SQL to use a hint to avoid the hash join OR
Use "hash_join_enabled = false" in your init.ora (a more drastic solution)

HTH --

Connor McDonald
BHP Information Technology
Perth, Western Australia
"The difference between me and a madman is that I am not mad"

Salvador Dali

Received on Sun Jan 10 1999 - 21:34:18 CST