Re: Index Compression question - solved, Thanks.

Jan

I suspect that your stats were up to date but that the process of swapping to compressed destroyed them. Did you drop and recreate the indexes or rebuild them ? If you rebuilt them did you specify compute statistics clause otherwise you loose stats (I believe) ?

Andy

"Jan Gelbrich" <j_gelbrich_at_westfalen-blatt.de> wrote in message news:b08okn$maie2$1_at_ID-152732.news.dfncis.de...
> Thanks Andy,
>
> Yes, we are using cbo, and that was why some tasks were much faster than
> others -
> I analyzed only a few tables and indexes after compressing, and only them
> were ok.
> All others were slower than before.
>
> I eventually analyzed one table with most complaints about it, and the
> problem was solved.
> I did not expect this to have *such* an effect, as the indexes were
analyzed
> just some days before,
> and I thought that there cant be much difference in between time ... but
> actually it has.
>
> Thatīs a lession for the day ... always use dbms_utility.analyze_schema( )
> after compressing indexes.
>
> Thank You
> Jan
>
> "Andy" <andy.spaven_at_eps-hq.co.uk> schrieb im Newsbeitrag
> news:U_QV9.3519$9R.12107447_at_newsr2.u-net.net...
> > Jan
> >
> > If you're using CBO, the stats for the indexes might need to be rebuilt.
> > That could explain why user stuff is slow (Oracle avoiding indexes
without
> > stats) and a heavy duty task (batch perhaps?) is running faster (it's
also
> > avoiding an index - this time one that wasn't a help - often the case if
> > you're reading a large %age of the table thru the index).
> >
> > Andy
> >
> >
> > "Jan Gelbrich" <j_gelbrich_at_westfalen-blatt.de> wrote in message
> > news:b08kg8$n044m$1_at_ID-152732.news.dfncis.de...
> > > Hello,
> > >
> > > due to most information I have (the doc, some books and NG FAQ),
> > > index compression can be good for performance,
> > > on compressable indexes only, of course.
> > >
> > > Last night I compressed half of my indexes (most composites, their
> > > cardinalities were good enough to do so),
> > > which reduced space requirements as I expected,
> > > but today, when all people are at their
> > > working places again, everybody complains that I made the whole system
> > slow
> > > down -
> > > expect of one heavy duty task, which runs as fast as never before.
> > >
> > > I know that there are various reasons why a system can be slow
> eventually,
> > > but for now I want to make sure that it was not due to my actions ...
> > >
> > > So my Question is: does index compression have *negative* performance
> > > effects ?
> > > If so, I would have to rebuild some crucial indexes again with
> NOCOMPRESS
> > > ...
> > >
> > > I would appreciate any comments from the experienced gurus ...
> > >
> > > Jan
> > >
> > >
> >
> >
>
>
Received on Fri Jan 17 2003 - 09:36:14 CST