Re: backup and recovery in data warehouse

Jeremy,

You say that you will not be able to back up all read/write tablespaces in a single cold backup, but also that you don't want to use archivelog mode. Given these two facts, how would you restore the database in the event it became necessary? Even if you restored separate pieces of the database from separate days, you would have no way to bring them consistent with each other. The restored database would be unopenable.

--
Jeremiah

jeremycdba_at_my-deja.com wrote:


> 

> Thanks for the replies. To answer a few of the questions raised.

> 

> This data warehouse will be at least 1.5 TB in size that's why we can't

> do nightly cold backups. Having some read-only tablespaces will reduce

> the amount to backup. But I expect that we'll never be able to backup

> the entire database in one go even excluding the read only tablespaces.

> Exports and imports will just be too slow.

> 

> We're thinking of NOT running in archive log mode. Being a data

> warehouse users will only be performing reads. The only updates/inserts

> will be performed during the loading of data, which can be reloaded if

> there are any problems. Having archiving turned off will improve the

> performance of this. So I can't see the point of running in archive log

> mode. Nice in theory I know so this may change in the future.

> 

> On the tablespace sizing and creation. We're going to be partitioning

> by month and each of these partitions will be in a seperate tablespace.

> So how do we know how big to make each tablespace and when should they

> be created? If we guess at the size and they're too small then we'll

> have problems, too big and we're wasting space. That's why I suggested

> using autoextend. And should I create them all for say two years before

> they're needed???