Re: Parallel Server and Parallel Query

A couple of people have suggested more information would be useful in providing some advice so here it is.

We are running IBM RS/6000 with AIX 4.2, HACMP 4.2, Oracle 7.3.3.

The servers each have 4 processors, and there are two servers in the cluster. Each server will run an instance. Available disk space on the RAID array is 32 Gb over 8 disks (after mirroring). We are going to be mirroring our data and apparently, AIX does not allow mirroring and OS level striping at the same time. The smallest we can create a logical volume is 8 Mb.

I hope this helps.

So, my questions are:

First Question:

When Oracle does a parallel query operation, how does it determine how to divide up the data between the query processors? Data location, data values? What? Strictly rowid? extent id?

Second Question:

When using parallel server option, you must use raw devices (is this true for AIX). You create a logical volume on one physical disk or across many physical disks. If you create a lv across many physical disks and use this lv as the datafile when creating the tablespace, does Oracle know how many physical disks the tablespace spans? If it does, then how does it know?

I ask this question because we are using the parallel server option and are considering using parallel query on some of our SQL. If Oracle can't detect the physical disks in a tablespace, then it might not be a good idea to use parallel query because it is really important to know where your data is sitting. Does anyone have experience with this?

Second Question, part 2:

Has anyone done any Oracle striping on a raw device before? I know that this could be very expensive (disk-wise), but we may need to do it anyway. I would appreciate any comments on how you did this and any other experience you had with this.

Doris Jetz
Consultant. Received on Fri Mar 07 1997 - 00:00:00 CST