RE: separate tablespaces for tables and indexes


I think the argument that is going to be made is that index io's and
table io's are sequential so there was never a performance benefit
within a single session regardless of disk configuration. =20

The second argument that would be made is that since IO is IO is IO, you
are better off figuring out which segments get the most IO and splitting
that across disks then blindly splitting all index and table segments.

There are some really smart people that argue there actually was a
benefit to splitting indexes and tables but they are in the minority,
most of the really smart people hold to the two arguments above.  As for
me I just try to figure out who the really smart people are and do what
they say :)

- Ethan

-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Mercadante, Thomas F
Sent: Monday, December 13, 2004 1:15 PM
To: 'JBECKSTROM@xxxxxxxxx'; oracle-l@xxxxxxxxxxxxx
Subject: RE: separate tablespaces for tables and indexes


In my view, separating indexes and tables was never about tablespace
contention.  It was always about moving the index and table files onto
separate disk drives to avoid disk contention.

Now that we have SAN disk where the 10 mount points that we see on a
box are actually split from the 3 hard disk drives on the SAN box, I'm
sure that splitting indexes and tables for performance matter anymore. =20

If you are going to use automatic allocation of segments for the
objects, it would be interesting to hear from the experts if the
of these objects really matters.

However, I still do it for management of database objects - mostly
index segment usage is so much smaller that table segment usage.

Hope this helps.



Other related posts: