(Originally posted 2005-03-09.)
PK01266 enables parallelism in Star Join under an additional set of circumstances:
From the APAR Database:
STAR JOIN QUERY ON NON-PARTITIONED FACT TABLE DOES NOT GET
CPU PARALLELISM CHOSEN WHEN MULTIPLE CPUS ARE AVAILABLE
Starjoin queries with non-partitioned fact table can encounter poor performance since CPU parallelism may not be enabled. The problem is corrected to enable CPU parallelism for this situation.
I suspect this will be most useful for SAP BW, but obviously other “Data Warehouse” users of Star Join should find it interesting as well.
An interesting paper in this area is Terry Purcell’s “Evolution of Star Join Optimization”.