How To Choose Best Stripe Size For A RAID10 Array - 10gR2

I have been studying possible ways to calculate the best size for the stripe size on my new storage. From what I have seen it appears that there is no straight answer but I believe there must be good way of estimating it. I am deploying a new storage that uses 16 x 600GB 15k RPM 6Gbps SAS drives, it is supposed to be very fast but I think if I get the stripe size on the RAID10 array right I should be able to get even more out of it. I am running oracle 10.2.0.4.0 (10gR2) on 64bit Linux with 32GB of RAM, the SGA is set to 20GB.

What I read said that the "stripe size must be at least as large as the I/O size." The I/O size is calculated from DB_BLOCK_SIZE * DB_ FILE_ MULTIBLOCK_ READ_COUNT, which is my case is "db_block_size = 16k * db_file_multiblock_read_count = 16 = 256k." The current RAID10 array that I am using which will be decommissioned soon has a stripe size of 128k, and if I understood it all correctly it is undersized for my Oracle instance.

With an I/O size of 256k the smallest stripe that I should choose is 256k but how do I know what is the best combination of db_file_ multiblock_ read_ count and db_block_size?

We are planning to create new table space with a block size of 32k as this has been recommended by a number of DBAs but with "db_block_size = 32k * db_file_multiblock_read_count = 16 = 512k".

Is this a viable size? Do you think this will actually improve performance on my system? Note that this database gets about 1500 insertions per second and peaks around 3000 insertions per second, it is a mixture of OLTP and data warehousing.

After doing a lot of research I came across an article by "Don Burleson" which says that starting in Oracle 10gR2 the "Oracle recommends not setting the db_file_multiblock_read_count parameter, allowing Oracle to empirically determine the optimal setting." References: URLs.....