Re: Clustering Tables

From: Jacob Love <jlove_at_engin.umich.edu>
Date: 1997/04/11
Message-ID: <5ilea4$s0o_at_srvr1.engin.umich.edu>#1/1


In article <3349AF0A.3CEC_at_compuserve.com>, R.Gauthier <robegaut_at_compuserve.com> wrote:
> I was wondering if anyone out there was using Oracle clusters and if
> so, what makes a table a good candidate for clustering ?

I assume you're talking about table clustering (as opposed to CPU clustering). Clustering saves physical space (under the right circumstances) and speeds queries. The trade-off is that inserts and updates become proportionately slower. Clustering was critically important in Oracle 5 and 6 because queries and reports against large tables (even well indexed ones) could be quite slow compared to similar sized decision support systems in competing (non-relational) products. The cost of disk space was an additional factor. With the dramatic improvements in speed and cost reductions provided both by hardware and software in the Oracle7 world, clustering becomes a harder case to make. In addition to the increase in overhead, clustering causes problems with database upgrades (at least in my experience, you're better off unclustering the tables before proceeding with any upgrade). Bottom line (imo): if performance issues do not mandate you to cluster by a fairly large margin, stay away from it.

-- 
-----------------------
Jack F. Love
Opinions expressed are mine alone, unless you happen to agree
Received on Fri Apr 11 1997 - 00:00:00 CEST

Original text of this message