Good point, but remember that speed isn't always the most important thing. Sometimes, uninterrupted availability can be a more important requirement... on 12/2/04 9:41 AM, Ken Payton at ken.payton@xxxxxxxxxxxxxxxxxx wrote: > Good point, should have touched that option. I haven't used it yet but > did realize it now exists. > > I would do some research before using this option on a large table > though. Depending on your requirements you may have to perform this > online and have no other choice. If the partition is say 10% of the > index size I would bet a invalidate and parallelized rebuild with lots > of sort area would accomplish the task much faster than a serial > exchange process. I'm assuming their is no way to parallelize the > deletes from the index. -- //www.freelists.org/webpage/oracle-l