Re: Creating datafiles for 500GB database

"Howard J. Rogers" <hjr_at_dizwell.com> wrote in message news:<opr6e5jlh33d8uqx_at_news.optusnet.com.au>...
> On 13 Apr 2004 16:34:06 -0700, Sandy N <urockblue96_at_yahoo.com> wrote:
>
> > Thanks for the thoughts and ideas (especially Dan, Howard, Mark). They
> > are very relevant to my pursuit :-). As far as my backup startegy, I
> > suggested using NOARCHIVELOG mode for the datamart. The reason being,
> > in case of any disaster, ALL the data can be recreated from the
> > mainframe. In any case we will take cold backup on every Sunday night.
> >
> > Regards
> > SN
>
> Ah well... that changes my advice then. Since your smallest unit of backup
> and recovery is the *database*, how you break that database up into data
> files becomes a matter of utter irrelevance.
>
> I favoured 2GB data files because in archivelog mode, that is then going
> to be the smallest possible recovery you can ordinarily do using O/S
> commands. But in noarchivelog mode, you're simply going to restore
> *everything* from last Sunday night's backup regardless. So you might as
> well go with the biggest data files your file system will permit.

I think you missed a critical bit there, Howard. As I read it, the Sunday night is a "nice-to-have," says nothing about what mainframe cycle is. Which could easily be "last night," in most situations I've seen. So from O's point of view, the unit is database, but from a reconstructionist :-) point of view it is probably which tables/referentials are most important rebuilt first.

This just points out that for Sandy, the decision on how big to make the datafiles lies elsewhere than backup/recovery. Whether that means partitioning is app dependent.

jg

--
@home.com is bogus.
"Bad Oracle." - hjr