RE: concepts document part about separating indexes and tablespaces

From: Pratap Singh (c) <>
Date: Mon, 19 May 2008 11:09:16 -0700
Message-ID: <>

This process is part of on going changes. I still remember the day when a 64GB striped/mirrored array was thrown at me when I was looking for 3 logical volumes for my data / index / logs. Yes I was adamant enough to reject the layout at first glance and took about a week to test and appreciate the performance. But on the other side it was 1995 if I remember correctly. Coming back to the point, what was once a ground rule is not applicable any more.
Multiple combination offered by SAN / Disk array makes data index separation obsolete.
Unless of course if you are supporting some old hardware/software with no RAID etc.  


PB Singh
DW Architect and Sr Data Modeler

From: [] On Behalf Of Mary Elizabeth McNeely
Sent: Sunday, May 18, 2008 9:57 AM
Subject: Re: concepts document part about separating indexes and tablespaces

I've gotten into this tiff with storage administrators at many a site.  

I have been informed by many of them that the "Oracle approved" disk layout is separate disks or separate disk pools for table and index access. (I could probably find an Oracle document recommending ANYTHING as the approved way of doing something if I looked long enough.) I've tried explaining that we are designing I/O for a multi-user system where I/O on various tables and their indexes would already be occurring concurrently (as Riyaj suggested earlier), so I didn't see much point in trying to separate the two. So, instead, why don't we "even out" the I/O among the disks by striping what would have been the table and index disks, together?  

I must be some unfortunate combination of not very eloquent, not very polite, and/or not very persuasive, because the person in charge almost always goes with what the storage administrators say, considering my idea heretical.  

The last time I had this tiff, I lost because the storage administrator was convinced that if the layout performed poorly, the manufacturer of the disk array might not support any performance related issues we had, because we had such a "non-standard" layout (striping tables and indexes in the same stripe/stripes). Sigh ...


  • Original Message ---- From: William Wagman <> To:;; Sent: Thursday, May 15, 2008 3:32:45 PM Subject: RE: concepts document part about separating indexes and tablespaces


I suspect this may date back to the "old days." I remember some years ago a statement, I believe it was in an Oracle Press book on oracle administration, that the ideal number of disks for a database was 28 (maybe not that many but close). This was in days when disks were smaller and performed differently. The idea was to spread things out across several disks, data, indexes, online redo logs, archive logs, undo, temp space, etc. so that I/O bottlenecks would be minimized. With today's faster and bigger (huger actually) disks this is not only less of an issue but difficult to do. How can you do that on a machine with 1 500GB disk? Why?  


Bill Wagman
Univ. of California at Davis
IET Campus Data Center
(530) 754-6208

From: [] On Behalf Of Bobak, Mark
Sent: Thursday, May 15, 2008 11:27 AM
To:; Subject: RE: concepts document part about separating indexes and tablspaces  

If it says that, someone should file a documentation bug to get it fixed.  



Mark J. Bobak
Senior Database Administrator, System & Product Technologies ProQuest
789 E. Eisenhower, Parkway, P.O. Box 1346 Ann Arbor MI 48106-1346
+1.734.997.4059 or +1.800.521.0600 x 4059 <> <> <>

ProQuest...Start here.  

From: [] On Behalf Of Rick Ricky
Sent: Thursday, May 15, 2008 2:17 PM
Subject: concepts document part about separating indexes and tablspaces  

i was talking to someone about this today. I cannot remember where in the Concepts document that it says that separating data from indexes improves performances?

-- Received on Mon May 19 2008 - 13:09:16 CDT

Original text of this message