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: 9i clustering Q

Re: 9i clustering Q

From: Ronald <devnull_at_ronr.nl>
Date: 15 Aug 2001 08:45:06 -0700
Message-ID: <67ce88e7.0108150745.5d3ba1a0@posting.google.com>


kw <none_at_nowhere.where> wrote in message news:<3B792A14.C6790E92_at_nowhere.where>...
> Hi all
>
> Is this possible to make query parallel on 9i RealApplicationCluster ?
>
> I mean if we have 3 Aix boxes (2x s80 and 1xSP14nodes)
> and we connect to one of them I want that query to be processed by every
> node I have (14) much faster
> then only by 1 node (even big one).
> Does it help when query will "touch" many partitions of one big table ?
>
> Is this possible and/or good idea to cluster such diffrent Hardware
> like IBM S80 and IBM SP ???

Hi,
parallel queries are very good possible on a RAC database since it can distribute the query over multiple instances (and hopefully multiple nodes) to do the work. I am not sure how the work is divided but I can imagine that splitting based on partitions is a first win that the database will go for.

As long as the HACMP and oracle versions on all members are identical you can run RAC on the cluster, even if they are on different hardware. If the performance is very different, it will be a bit harder to divide the workload evenly distributed in time. A table is split in parts to be divided over the different instances in a PQ. It is very possible that an instance on a very slow node receives a part as big as one on a very fast node. The merge of results can not complete before the last instance is finished. Maybe you can compensate the performance difference by having more instances of the same db on the fast nodes and less on the slower nodes.

Ronald.



http://www.ronr.nl/unix-dba Received on Wed Aug 15 2001 - 10:45:06 CDT

Original text of this message

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