Re: Parallel query question
From: Robert Fazio <rfazio_at_home.com.nospam>
Date: Wed, 15 Aug 2001 01:24:29 GMT
Message-ID: <hbke7.104188$EP6.26789758_at_news1.rdc2.pa.home.com>
Date: Wed, 15 Aug 2001 01:24:29 GMT
Message-ID: <hbke7.104188$EP6.26789758_at_news1.rdc2.pa.home.com>
If I understand the question right. Only consider the instance that has the tables that you are selecting from. That is where the work will be done. Not on the one where you are running. Why not rely on the table's default parallel clause rather than making the assumption of what you should do. Most DBA's will define the Parallel clause for a table to be the best for that table. The number of threads should be more dependent on the amount of data being processed than on the number of processors in the system.
-- Robert Fazio Senior Technical Analyst dbabob_at_yahoo.com "bob sullivan" <bsullivn_at_home.com> wrote in message news:3B789CCF.C585D91_at_home.com...Received on Wed Aug 15 2001 - 03:24:29 CEST
> We've got an 8.0.5 instance (call it instance A) and a 7.3.4
> instance (call it instance B) connected by a db link, and I've
> got a stored procedure in A that selects data from B and inserts
> it into A. Instance A is on a server that has 4 processors,
> and instance B is on a server that has 2 processors. Both
> instances have the Parallel Query Option enabled, and I'd like to
> try to take advantage of that fact in my stored procedure.
>
> I've always used the rule of thumb that the degree of parallism
> in a 'parallel()' hint should be 1.5 times the number of
> processors. Since the code is stored in instance A, but selecting
> data from instance B, should my hint have a degree of parallelism of
> 1.5 * instance_A_processors == 1.5 * 4 = 6, or should I use a
> degree of parallelism of 1.5 * instance_B_processors ==
> 1.5 * 2 == 3?
>
> Thanks in advance for the help,
> bob
> bsullivn_at_home.com