RE: standby database
- From: "Mark W. Farnham" <mwf@xxxxxxxx>
- To: <dbvision@xxxxxxxxxxxx>
- Date: Fri, 6 Mar 2009 11:20:42 -0500
At least Oracle 6.0.33 for large smp.
Prior to 6.0.33 there were too frequently broken redolog files due to race
conditions not quite trapped for this to be stable. Probably it was okay
sooner on fewer cpus. I only had trouble on 20+ cpu Sequents. Had to keep
track of what would require a reinstantiation of the standby for yourself.
Getting a read only frozen version required cancelling recovery, shutting
down, copying the whole thing (locally or to yet another server). Then if
your copy was local, a startup rename was required so that recovery on the
standby could be resumed. The nice thing about that was the opportunity to
augment the frozen-renamed standby copy with additional schema objects and
aggregations and cleansing to feed a warehouse or run otherwise expensive
reports. Since the data was frozen aggregates didn't have to be maintained
and could be created before the users were turned loose so the consistent
read load was dramatically less than the "real" system. Then at "user time"
all the data could be static except temp for sorting and groupbys and block
cleanouts. Accountants also loved having no moving targets and would think
twice about getting exactly up to date reports on the "real" system because
of the speed difference. Creating a load shift onto the standby machines
also was often the convincing factor that tipped the balance to getting
enough horsepower at the remote site to actually run production if you had
to fail over. Far too many configurations to this day claim to be business
continuation sites rather than merely remote data recovery sites, mostly due
to lack of horsepower on the "business continuation" site. (I have no
problem with folks running standby recovery on much less than production
capable hardware as long as they understand what they have and don't
overstate capability.)
Oracle 7.3+ for "supported standby"
IOUW '95 paper by Lynne Thieme was essentially the product announcement
(that being the last IOUW that was still Oracle's primary marketing event as
well as the IOUG's event.)
The only quibble I would make with what you wrote is that it was probably
LESS buggy in its original form on 7.3, because it was less ambitious and
put more of the essential steps on the user, together with documenting what
you could not do as opposed to trying to prevent you from doing it in an
automated way. And of course it was not called "Dataguard" yet.
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx]
On Behalf Of Nuno Souto
Sent: Tuesday, March 03, 2009 4:27 AM
Cc: oracle-l@xxxxxxxxxxxxx
Subject: Re: standby database
Niall Litchfield wrote,on my timestamp of 3/03/2009 5:48 PM:
> can anyone tell me when this first came out - I'm guessing between v7 in
> 1992 and v8 in 1997, probably in 7.3.4 but would really like a
> definitive answer
I was doing it with v6 in 1990. ;) Anyways,
from a vague memory, I'd say 8 and 1997. It might have been
announced somewhere in the long v7 lifecycle but it'd have
been impossibly buggy to use then. Still is in some platforms,
but that's a different story...
--
Cheers
Nuno Souto
in sunny Sydney, Australia
dbvision@xxxxxxxxxxxx
--
//www.freelists.org/webpage/oracle-l
--
//www.freelists.org/webpage/oracle-l
Other related posts:
