Re: Building indexes in parallel

From: Martin Douglas <dafis_at_ppp.test.jccbi.gov>
Date: 1996/09/28
Message-ID: <NEWTNews.843938775.5263.dafis_at_ppp.test.jccbi.gov>#1/1


In article <324B287C.470F_at_ix.netcom.com>, <wmuntz_at_ix.netcom.com> writes:
> Path: news.ou.edu!news.ecn.uoknor.edu!solace!news.stealth.net!www.nntp.primenet.com!nntp.primenet.com!howland.erols.net!netnews.com!ix.netcom.com!news
> From: Walter Muntzenberger <wmuntz_at_ix.netcom.com>
> Newsgroups: comp.databases.oracle
> Subject: Re: Building indexes in parallel
> Date: Thu, 26 Sep 1996 20:06:05 -0500
> Organization: QED Systems Consulting
> Lines: 33
> Message-ID: <324B287C.470F_at_ix.netcom.com>
> References: <3242a297.6677726_at_n5.gbso.net> <51v5dt$et7_at_newsa.netnews.att.com>
> Reply-To: wmuntz_at_ix.netcom.com
> NNTP-Posting-Host: ara-il7-22.ix.netcom.com
> Mime-Version: 1.0
> Content-Type: text/plain; charset=us-ascii
> Content-Transfer-Encoding: 7bit
> X-NETCOM-Date: Thu Sep 26 6:05:23 PM PDT 1996
> X-Mailer: Mozilla 3.0 (Win95; I)
> To: Naren Chintala <naren_at_mink.att.com>
>
> Naren Chintala wrote:
> >
> > chuckh_at_dvol.com (Chuck Hamilton) wrote:
> > >Is it possible to build a primary key index in parallel? If so, what's
> > >the syntax?
> > >--
> > >Chuck Hamilton
> > >chuckh_at_dvol.com
> >
> > Don't do it. It will fragment you index. Email me if you need more information.
> >
> > Naren Chintala
> > naren_at_mink.att.com
>
> What do you mean by fragment your index?
>
> --
> _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
> _/
> _/
> _/ Walter Muntzenberger _/_/ _/_/_/ _/_/_/
> _/
> _/ Principal Consultant _/ _/ _/ _/ _/
> _/
> _/ QED Systems Consulting _/ _/ _/_/_/ _/ _/
> _/
> _/ Email: wmuntz_at_ix.netcom.com _/ _/ _/ _/ _/
> _/
> _/ Text pager: 800.SKY.TEL2 pin 1645838 _/_/ _/_/_/ _/_/_/
> _/
> _/ _/
> _/
> _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
>

Fragmentaion should not be a problem if you manage it properly. When you have multiple hard drives that span an index tablespace and you use manual stiping You can set you degree of parallelism to the number of devices. If you make you initial and next extents the same size as you datafile, then performance should actually increase. With the new hashing algorithims in 7.3 databases as well as parralelism, it can benefit you to startoff in extents. Received on Sat Sep 28 1996 - 00:00:00 CEST

Original text of this message