Re: Why is on-line backup one tablespace a time?

You do it one tablespace at a time for one simple reason: putting a tablespace into hot backup mode causes Oracle to begin generating BLOCK-sized redo for every transaction that hits that tablespace.

Ordinarily, if you do an 'update emp set sal=900 where name='Bob', then the redo generated would be a small transaction number, the old value in the sal column, and the new, plus some bits and pieces about rollback segments being used. In other words, for your typical update, we are talking byte-sized redo.

That same transaction, done whilst in hot backup mode, will now generate an entire 8K (or whatever your block size is). Kilobyte-sized redo.

Since many people's performance issues are directly related to a sluggish redo sub-system in the first place, the idea of swamping it further with a thousand times the quantity of redo, across the entire database, doesn't bear thinking about.

Hence: one at a time, and you keep the time spent between 'begin backup' and 'end backup' as short as possible before moving on to the next one.

Provided you retain all redo from the time the *first* tablespace went into hot backup mode, the fact that your backup copies are not from the same timeslot is irrelevant for the purposes of recovery.

Regards
HJR "Dino Hsu" <dino1_at_ms1.hinet.net> wrote in message news:dj3ioto46anmoilhuh6h00mijshio6jhf3_at_4ax.com...
> Dear all,
>
> I read chapter 12 Overview of Backup and Reovery of Jason Couchman. An
> on-line backup is performed with one tablespace a time with:
> alter tablespace <ts> begin backup;
> make backup of datafiles with OS copy command
> alter tablespace <ts> end backup;
> These tablespace backups are not required to happen at the same time,
> and the database is still open. I wonder how these tablespace backups
> can be synchronized. Anyone comment on this? Thanks in advance.
>
> Dino
>
Received on Sun Aug 26 2001 - 15:00:52 CDT