Re: Backup and recovery of Large databases

Carl_at_carlc.demon.co.uk (Carl Churchman) writes:

>I am the Senior System administrator/DBA of a Sequent SE60.We
>have a large database of 40Gb which has to be highly available.
>At present we run hot backups 6 nights a week and run archiving.
>A cold backup is run on a Saturday night.These backups take about
>4 hours (not bad eh?) using 4 * 10GB (compressed) Exabyte tape drives
>and PTX/Backup. My problem is this:
> I have tried to recover the database on another machine,to test
>disaster recovery.The restore is still running and by my calculations
>is going to take about 42 hours!.How are you other DBA's backing up and
>recovering your large databases and how long do they take?
>

Carl, this is very common in large SMP shops where >1 CPU does the transactions but 1 CPU does the recovery. The GOOD NEWS is that Oracle7.1 allows you to specify a degree of parallelism with the recover database/tablespace/file command. The current version on Sequent is 7.1.3.

You can now choose how many CPUs are rolling forward a database or you could choose 1 for a low transaction rate tablespace and 8 (at the same time) for your high transaction rate tablespace. You would have to work out the combination that maxes out all your CPUs and I/O in a balanced manner.

Regards,

-=:[<{()}>]:=--=:[<{()}>]:=--=:[<{()}>]:=--=:[<{()}>]:=--=:[<{()}>]:=--=:[<{()} David Natelson Oracle Corporation dnatelso_at_oracle.com -=:[<{()}>]:=--=:[<{()}>]:=--=:[<{()}>]:=--=:[<{()}>]:=--=:[<{()}>]:=--=:[<{()} Received on Tue Oct 11 1994 - 02:09:26 CET