Looking for opinions on building IX's after loading data

We have a data warehouse that contains large tables and matching large indexes.

There is a small debate being waged amongst the DBAs here regarding the best way to handle the
indexes.

All of the tables are being populated using Sql*Loader with the direct-path option and "unrecoverable."

One side argues for the "drop the index/load the table/rebuild the index" strategy.
This avoids the "direct path load state" problems and allows for relatively quick rebuilds using
parallel index creation.

The other side argues for allowing Sql*Loader to do its thing; merging new index entries into
the old index segment. The argument given here is to avoid creating new processes for the DBAs to maintain.

Does anybody have any performance benchmarks measuring the difference between the 2 options?
Anybody have any opinions regarding which method is best?

Thanks in advance. Received on Mon Oct 30 2000 - 16:00:09 CST