Need Cluster Help

Greetings,

We are running Oracle 7.2.3 on HP-UX 9.4.

We have a 15G table which experiences very large batch inserts and deletes. Delete performance is very poor. The index seems to grow excessively.

It has been suggested that we can improve delete performance and index efficiency by creating a cluster which includes only this table. The cluster key would be the two columns by which we do the delete.

The suggested cluster key is very non-discriminating. We might have a million rows in the table and only 60 distinct values for the suggested key. The distribution of rowids across key values is uneven.

===

My understanding is that clusters are intended to improve join performance. A cluster of one table does not achieve that. I also understand that a cluster key should be selected such that all the rowids for a given key value fit in a single block. The suggested key is not selective enough to do that.

===

Does anybody have any experience using clusters? Can they help batch insert and delete operations? Improve index efficiency? Do they cause additional maintenance problems? Does a cluster of one table make sense?

Any comments at all are appreciated.

Thanks,

Jim Peterson

-- 



=============================



Jim Peterson
TDS Computing Services
Madison, WI USA
james.peterson_at_teldta.com