Stripe and Mirror Everything - The "S.A.M.E. Method"

Here's the url for a white paper titled "Configuring the Oracle Database with VERITAS Software and EMC Storage for Optimal Scalability, Manageability, and Performance" on Technet:

http://technet.oracle.com/deploy/availability/pdf/ora_cbook1.pdf

The conclusion of the paper is something like "just put everything on RAID0+1 and don't worry about file contention by using the 'SAME' method." SAME is an acronym for "stripe and mirror everything." They compare the old "traditional method" of segregating redo logs, RBS, temp, data, index, and system tablespaces/datafiles to the SAME method with everything residing on one large array. They give benchmarks to show that the SAME method actually performs better.

Any feedback or criticism of this paper? Does it really reflect the way we should configure very large databases? Their test database was 50G.

I'm curious about the validity of the benchmarks. It seems their implementation of the "traditional method" is half hearted. The only files they separate are the redo logs. What if they kept the same array as in test case 2 and duplexed the redo logs on drives outside the array? What if they went further and implemented physical drive separation with three volumes: one for data; one for indexes; and one for system, RBS, and temp tablespaces? What if they had a 500G database? Another curious observation is that db_block_size was only 4K on a Solaris E3000.

The paper says, "Today, most Oracle DBAs do not configure an Oracle database and storage based on the SAME method because that is not how they were trained to implement Oracle and storage." It goes on to say, "...with the right storage software and hardware solution, we recommend that Oracle DBAs consider the new SAME approach." Hmmm... are they trying to tell us old DBAs to keep up with the technology?

All opinions, humble and otherwise, are earnestly requested. Received on Fri Jan 12 2001 - 11:54:40 CST