Clusters and insert performance

Hi

I'm on 8.0.5 (I guess...., the jury is still out....) and have some question about clusters.

We have a master - detail1 - detail2 setup of tables, and every hour we need to insert a huge(?) set of data into the detail1 - detail2 tables.

Detail2 will always be accessed with detail 1. Detail 1 is never interesting with at least on detail 2 row. The ration between size of detail 1 and detail 2 is about 1:100, both for rows and bytes.(Both tables are narrow, about 25 bytes) I have the whole set of data for detail 2 for a given detail 1 available togheter before insert.
Updates will not happen to detail 1 nor detail 2. Deletion happens periodically of detail 1 rows with all corresponding detail 2.

Clustering detail 1 and detail 2 will obviously(?) give a performance gain for queries, but all documentation advises me to be careful about insert performance. Will the given sequential nature of my data reduce the impact?
Can I insert/load the data in some creative way exploiting the sequential nature, and get some performance gain?

How about partitioning, given different disk, the inserts should really speed up, cluster or no cluster? I think partitioning on master is a viable option, which should give insert and query performance gain.

Please excuse my non-specific questions, but it's all I have to go on so far. Any ideas, views, pointers or hints are welcome. Do not hesitate to make the necessarry assumptions to give a useful answer.

Thanks!

ole c Received on Fri Sep 17 1999 - 07:23:18 CDT