Re: Creating datafiles for 500GB database

From: Mark Bole <makbo_at_pacbell.net>
Date: Wed, 14 Apr 2004 01:40:21 GMT
Message-ID: <9E0fc.36637$PE.24184@newssvr29.news.prodigy.com>

Howard J. Rogers wrote:

> On Tue, 13 Apr 2004 02:23:40 GMT, Mark Bole <makbo_at_pacbell.net> wrote:
>

>> Howard J. Rogers wrote:
>>
>>> On 11 Apr 2004 18:45:34 -0700, Sandy N <urockblue96_at_yahoo.com> wrote:
>>>

[...]

>>>
>>> Generally, the unit of backup and recovery is the data file, so to my 
>>> way of thinking 10GB is too big. I prefer no bigger than 2GB. And 250 
>>> data files would not be outrageous for a database (though you'd want 
>>> to make sure you didn't let the default MAXDATAFILES control file 
>>> sizing parameter kick in: it's 30!
>>>
>>> Regards
>>> HJR
>>
>>
>> I think of the unit of (physical) backup and recovery as the tablespace.

>
> Well, you're welcome to think anything you want of course. But the fact
> remains that the minimum unit of backup is the data file. Particularly
> since a tablespace is a component of Oracle's *logical* storage model,
> not its physical.
>
> The minimum unit of recovery in 9i, of course, and using RMAN is the
> data block, but otherwise, it's the data file too.
>
> Those are just facts, I'm afraid.
>
> And it's why I wouldn't be too happy to have to restore a >5GB data file
> to perform the smallest possible recovery using O/S techniques.
>
> But YMMV, so that's fine.
>
> Regards
> HJR
You're right. I was really just trying, in a sloppy way, to bring in a related concept: it's not just the size of the datafiles but the make-up of the tablespaces that should be considered (for example, the partitioning option mentioned by another poster is more useful if tablespaces are configured with partitioning in mind).

According to Oracle, the only "logical" backup is an export file. I have been conditioned by the "ALTER TABLESPACE ... [BEGIN | END] BACKUP" statement to think of tablespaces as a unit of physical backup.

--Mark Bole Received on Tue Apr 13 2004 - 20:40:21 CDT