Re: fragmentation and index maintainence

drak0nian_at_yahoo.com (Paul Drake) wrote in message news:<1ac7c7b3.0405181928.6e5ef427_at_posting.google.com>...
> Jeremy <newspostings_at_hazelweb.co.uk> wrote in message news:<MPG.1b1439c81cf83992989bf5_at_news.individual.net>...
> > In article <c86ce4f.0405180705.6bc45f56_at_posting.google.com>,
> > DiggidyMack69 says...
> > > "Howard J. Rogers" <hjr_at_dizwell.com> wrote in message news:<4092660b$0$4543$afc38c87_at_news.optusnet.com.au>...
> > > > DiggidyMack69 wrote:
> > > >
> >
> > <snip>
> >
> > > > >
> > > > > What are good rules of thumb to look at in regards to when looking for
> > > > > signs of fragmentation and other problems of the like?
> > > >
> > > > A good rule of thumb?
> > > >
> > > > Don't bother.
> > > >
> >
> > <snip howard's suggestions>
> >
> > >
> > > Thanks for the info. For the record I was talking about index
> > > fragmentation. It has been told to me that some of my indexes are
> > > 'fragmented' by an 'authority'. That is all. I didn't have any stats
> > > or anything to back-up or disregard that claim. Thanks to folks here
> > > I have alot more useful information. I assume that some automated
> > > tool has been run to generate a report of 'fragmented' indexes.
> > >
> > > DM
> > >
> > Intermedia Text indexes need regular rebuilds otherwise the query
> > performance can be severely degraded.
>
> Jeremy,
>
> that's not a rebuild in terms of index fragmentation, the fact that
> explicit maintenance is required is due to the fact that intermedia
> context indexes are not maintained synchronously, so the become
> "STALE", not fragmented.
>
> Oracle maintains b*tree and bitmap indexes synchronously, and
> recursively.
>
> There are 2 methods possible for synching the context indexes,
> internally via a dbms_job calling a procedure, or externally via the
> ctxsvr executable.
>
> IIRC, Daniel Morgan has posted in the past that he prefers the
> internal method due to some conditions that can cause the indexes to
> become larger via the external routine. Maybe he'll refresh us as to
> what he specifically mentioned.
>
> Pd

Sybrand,

I was not attempting to spread FUD.

Daniel,

Sorry that my memory was not quite correct, it was Connor that posted it to c.d.o.s. If Connor posted that he has benchmarks, that is good enough for me.

Paul

Here is the actual text:
http://www.mailarchive.ca/lists/comp.databases.oracle.server/2004-01/0351.html

<block quote>
"as well as what others have said, I can add...

DO NOT USE CTXSRV !!! Benchmark tests we've done show that you can balloon the size of an intermedia by a massive scale as opposed to using the sync_index routines.

ctxsrv is now deprecated by oracle.

hth
connor
</block quote> Received on Fri May 21 2004 - 13:47:12 CDT