If it were an option, I would load into a staging table, build indexes, and partition exchange. This yields the best of all worlds: parallel direct load with no index maintenance, parallel nologging index builds, and doesn't effect queries. My recommendation: Leverage ASM and not worry about it. You will get the best performance with the least effort. On 2/8/08, ryan_gaffuri@xxxxxxxxxxx <ryan_gaffuri@xxxxxxxxxxx> wrote: > Are there any benchmarks testing whether there are performance improvements > on reads and writes when you separate data and indexes on separate disks? > > cases: > > 1. sqlloader direct path load of a large number of records to a table with > 4-5 indexes. The indexes can't be dropped during loads since queries take > place during the load. > > My understanding is that the write process is serial so separating the data > and indexes onto separate disks will not improve performance much. I don't > have the disks to test this on right now. > > 2. queries. mainly oltp type queries. any performance improvements for > separating the hard disks that tables and indexes are on? > > This is on a SAN, but we can have it configured to give us separate disks and > show them to us as separate mount points. -- Regards, Greg Rahn http://structureddata.org -- //www.freelists.org/webpage/oracle-l