Re: Online creation/extension of Tablespaces possible??

From: Howard J. Rogers <hjr_at_dizwell.com>
Date: Sun, 12 Sep 2004 12:09:02 +1000
Message-ID: <4143afc4$0$18851$afc38c87@news.optusnet.com.au>

Daniel Morgan wrote:

> Sybrand Bakker wrote:
> 

>> On Sun, 12 Sep 2004 06:44:47 +1000, "Howard J. Rogers"
>> <hjr_at_dizwell.com> wrote:
>>
>>

>>>Too many data files in a database are themselves not a terribly good idea
>>>for efficient checkpoints etc, so I wouldn't want to see 2000 files of
>>>10MB each to make a 20GB database, either.

>>
>>
>> Which was exactly the context I was implicitly referring to.
>> I don't think there is much disagreement between us, if I can, I would
>> limit my datafiles to 2G
>>
>>
>> --
>> Sybrand Bakker, Senior Oracle DBA

> 
> How does this correspond with using an uncooked file system? Surely you
> wouldn't break up a 70GB drive into 35 separate partitions? So why the
> concern about file size?

An "uncooked file system" is a contradiction in terms, surely? Either one has a file system, or one has raw ("uncooked") partitions. But not both.

And yes, I may well break a large hard disk into many smaller partitions. If I was using raw, I would probably be talking SCSI rather than IDE, so my disk size is likely to be in the 36GB range rather than 70 or 80GB. So the problem is not quite so drastic as you make out.

But yes, the concern about file size is one of backup and recovery (for me, anyway... can't speak for Sybrand). And that doesn't go away as an issue just because you're using raw.

You are aware, too, I'm sure that Oracle's RAC courseware has a slide in it where it carefully instructs users of raw devices to create something like a *minimum* of 16 separate partitions to house the cluster database. So it happens.

Regards
HJR Received on Sat Sep 11 2004 - 21:09:02 CDT