You seem to have missed the words "for each of the datafiles" in the statement that "striped logical volumes can be used". The problem with most striped configurations is not the striping - that's actually an advantage; it is the sharing of physical disks between related stripes that causes the disk heads to thrash when they are read in parallel. If each datafile is striped over a set of dedicated disks then you get the best of both worlds.

EMC disk, its inherent lack of striping and what to do about it is treated on the Ixora web site in the tip on Disk Configuration for Pseudo-Random Reads. Your partitions are too small to give you much flexibility. You can either follow the EMC configuration ideas presented in that tip and keep the degree of parallelism down to say 4, or raise the degree of parallelism to about 12 and manage without striping.

I have a small doubt regarding parallel processing on tables stored on EMC disks. I read your paper about disk configuration where you have mentioned the disadvantages of using striped disks for parallel tablescans and have a few doubts regarding that --

a) In the "Why not striped disks section" you have mentioned that striping is not suited for parallel table scans and in the "What is the alternative" section you mention that striped logical volumes can be used for large tables to improve the transfer rates. I find this slightly confusing, can you please enlighten me on this?

b) I have a couple of large tables (around 10 Gigs) with a few huge indexes on them, these tables are partitioned on a monthly basis (about 4 years worth data) and the main activities on these tables are insertions and report generation. These tables are currently stored on EMC disks and I was wondering if the points you have mentioned in your paper about parallel scans would be relevant to these PARTITIONED tables on EMC disks as well.