In the early 1980's, researchers recognized that semantic information stored in databases as integrity constraints could be used for query optimization. A new set of techniques called semantic query optimization (SQO) was developed. Some of the ideas developed for SQO have been used commercially, but to the best of our knowledge, no extensive, commercial implementations of SQO exist today.

In this paper, we present an implementation of two SQO techniques, Predicate Introduction and Join Elimination, in DB2 Universal Database. We present the implemented algorithms and performance results using the TPCD and APB-1 OLAP benchmarks. Our results show that SQO can lead to dramatic query performance improvements. A crucial aspect of our implementation of SQO is the fact that it does not rely on complex integrity constraints (as many previous SQO techniques did); we use only referential integrity constraints and check constraints.