Re: Tablespace distribution on 3 external hard disks

"Bib Endum" <gd-newsgroups_at_spamex.com> wrote in message news:R9m6c.21663$E71.1527005_at_news20.bellglobal.com...
>
> "Howard J. Rogers" <hjr_at_dizwell.com> a écrit dans le message de
> news:40591262$0$31904$afc38c87_at_news.optusnet.com.au...
> >
> [snip]
> > checkpointing) redo allocation retries (and so on). It would be better
> under
> > those circumstances to create table 1 with three extents and arrange for
> > each extent to be on a different hard disk (so that's a tablespace
> spanning
> > all three disks). The other two tables could go on pretty much any hard
> disk
> > (so one more tablespace spanning all three disks might be the go).
>
> I always wondered what oracle does in this case: Suppose you have an empty
> tablespace of 3 Gb divided in 3 datafiles of 1 Gb each on 3 different
> drives.
>
> You start inserting like mad in that tablespace. Will oracle be
intelligent
> enough to insert in each databafile in a round robin manner or will it
fill
> the first datafile, then the second, then the third ? Or randomly pick a
> datafile ?
>
> The basic question is: Is oracle aware of the disk layout in some way.
Yeah
> I know it's "aware", it's in the dictionnary, but does the program care
> about that ?
>
> Otherwise you could use OS utilities to stipe a volume across the 3 drives
> and let the OS do the load balancing. Which is best ?

See the "Poor Man's RAID" thread elsewhere in this newsgroup. The answer is that Oracle will indeed stripe across all three disks, although the behaviour doesn't kick in for locally managed tablespace until the first 1MB of extents has been allocated.

Which is best? Hardware raid has to be, of course. But getting Oracle to do it for free in the absence of anything else is a bit of a bonus.

I don't think I would ever touch O/S software RAIDing: the O/S should have better things to be doing.

Regards
HJR
> Thanks for your eventual clarification.
>
>
Received on Thu Mar 18 2004 - 15:22:00 CST