Multiple data files per tablespace--benefit?

Traditionally I've used multiple data files for large tablespaces. Manageability benefits aside, when using traditional filesystems, performance on all sorts of operations (typically parallelized) has been much better if I had separate data files on separate disk devices.

Yesterday I did some benchmarking using ASM instead of a filesystem, and letting ASM stripe across disk groups. I saw no advantage whatsoever in having multiple datafiles. In fact, for parallel bulk operations (nologging bulk inserts, full scans, etc.) performance degraded quite a bit.

Just wanted to confirm, before making any wholesale changes, that from a performance perspective it is reasonable to have single, large data files in an ASM world?

Bob Received on Thu Jan 05 2006 - 16:34:12 CST