Re: Oracle 7 parallel Server

From: Brian P. Mac Lean <brian.maclean_at_teldta.com>
Date: 1996/07/31
Message-ID: <31FF808F.2A5E_at_teldta.com>#1/1


Gert Struyven wrote:
>
> Hi,
>
> We're planning to work with Oracle 7.3 parallel server on an IBM J30
> bidirectional server.
>
> Because this will be the first database at our site running in parallel
> mode, we haven't got much experience.
>
> Is ther someone that already implemented parallel server and is willing to
> tell me some pittfals, good advices, etc...
>
> I'dd be gratefull,
>
> Gert Struyven
> Head DBA-departement
> ABP
> The Netherlands

At the risk of being spamed, I'll give this a try.

I have been to more than a few Oracle Education classes to include "Oracle Parallel Server". I feel that it is the most complicated Oracle environment I have seen yet. The challenge is to prevent block swapping between machines, or what Oracle calls "pinging". The details can only be fully described and understood by taking the class. But in short, you need to partition the data so that requests for (the same) database blocks don't occur from more than one machine. The easiest way to do this is run different application systems, that have few overlaps, on different machines. Say inventory on machine one, and payroll on machine two. If you require access to the same table data from more than one machine you need to look into the use of "free list groups" (it's part of the create table statement), and this is where it gets way to complicated for this small e-mail.

The only reason I would ever be a willing participant in a Parallel Server environment is if the need for "high availability" existed. To do this just to get more cpu cycles is crazy. If more cycles is what you want, buy into the Sun Solaris product line. I believe it's scaleable to 50+ processors (if not now, soon).

Good luck and you have my sympathies.

brian.maclean_at_teldta.com Received on Wed Jul 31 1996 - 00:00:00 CEST

Original text of this message