Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: extended run time with parallel query on.

Re: extended run time with parallel query on.

From: Jonathan Lewis <jonathan_at_jlcomp.demon.co.uk>
Date: Tue, 30 Mar 1999 21:17:27 +0100
Message-ID: <922826301.27903.1.nnrp-09.9e984b29@news.demon.co.uk>


I note you have INSTANCE 2 - are you really running a cluster with 2 instances of Parallel Server.

In general, the answer to your question is yes, I have seen parallel queries running slower than serial queries. PQ usage encourage parallel tablescans. If you have not planned for them then:

  1. you can easily get massive I/O contention
  2. you can get massively sub-optimal paths

In some cases you can find that the parallelism is very badly skewed, but this tends to be quite rare.

Check the plans, I/O figures, and v$pq_tqstat.

There are some articles on my web-site which may be of interest.

--

Jonathan Lewis
Yet another Oracle-related web site: www.jlcomp.demon.co.uk

akudirka_at_my-dejanews.com wrote in message <7do5e2$9e7$1_at_nnrp1.dejanews.com>...
>I have several batch jobs where their run times increased by as much as
400%
>when I made tables eligible for query parallelism. Has any one seen that
>same behavior and found a solution?
>Sun sparc 2000 2.5.1 with 16 processors, 5 gig rollbacks, gig memory,
>degree 4 - instance 2. Some of the queries are doing table scans.
>Avg. table size 2.5 gig. Oracle 7.3.4
>Init settings = engueue resource 5000, processes 1500
Received on Tue Mar 30 1999 - 14:17:27 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US