Comments at end.
HJR
--
Resources for Oracle: http://www.hjrdba.com
===============================
"Martin" <mcantwell_at_paradigmone.com.au> wrote in message
news:d094c395.0111282152.2bf45c10_at_posting.google.com...
> "Howard J. Rogers" <howardjr_at_www.com> wrote in message
news:<3be10e0b$0$9813$afc38c87_at_news.optusnet.com.au>...
> > Comments embedded.
> > Regards
> > HJR
> > --
> >
> > Oracle Resources : http://www.geocities.com/howardjr2000
> > ========================================
> >
> >
> > "Prem Alleppey" <prem_alleppey_at_yahoo.com> wrote in message
> > news:bf7b9060.0110311411.560493bc_at_posting.google.com...
> > > We are looking into using Oracle Standby for disaster recovery for an
> > > OLTP database and would like to keep the potential data loss to be
> > > about 5 minutes worth.
> > >
> > > A couple of questions in this regard, especially from those who use
> > > Standby for DR:
> > > * If we force log switching to happen every 5 minutes, what would be
> > > the performance impact of that?
> >
> > Disastrous.
> >
> >
> > >Would it be significant or negligible?
> >
> > Significant
> >
> > > What would could we do to minimize the impact?
> >
> > Nothing. Just don't go there, OK?
> >
> > > * How much time does the ARCH process take to archive the files once
> > > the switch happens? What can we do to make sure that ARCH is not
> > > lagging behind?
> > >
> >
> > Have a billion and one online redo logs? Look -it's a *really*,
*really*
> > bad idea. Standby databases give you a *level* of security, but not
that
> > sort of level! If you want that degree of protection from data loss,
> > consider Parallel Server or replication strategies.
> >
> > Better yet, upgrade to 9i, where Data Guard really can give you the 5
minute
> > level of protection you're after, policed, resourced and with
potentially
> > huge impacts upon the availability of the production database.
> >
> > Don't do it. Please!
> >
> > Standby databases are designed to be the current redo log behind the
> > production database. You start adjusting your redo log sizes to
translate
> > that into a 5 minute lag, and you are looking disaster squarely in the
eye.
> > Checkpoints are awful affairs for performance. You'll be having one
every 5
> > minutes. It doesn't bear thinking about.
> >
> > Regards
> > HJR
> >
> >
> > > Thanks in advance..
>
> I do not agree, we run this on 9i under Solaris on a one minute lag,
> without experiencing performance issues. If you actually test this you
> will see very little difference in performance, there is some but not
> much.
Very different kettle of fish. 9i is designed to support no data
divergence/no data loss models (I'm assuming you are running data guard,
which is one of the things I recommended). 8i and below isn't. The
original poster was talking about reducing the size of his redo logs to
arrange for switching at a rate to give him a satisfactory lag. 9i switches
at a normal rate and introduces the lag via a completely different
mechanism. Since you are switching at a normal rate, I wouldn't expect to
see any performance hassles.
Regards
HJR
Received on Thu Nov 29 2001 - 05:32:59 CST
