WPI - CS - TR - 99 - 06 March 1999

More than ever before schema transformation is a prevalent problem that needs to be addressed to accomplish for example the migration of legacy systems to the newer OODB systems, the generation of structured web pages from data in database systems, or the integration of systems with diierent native data models. Such schema transformations are typically composed of a sequence of schema evolution operations. The execution of such sequences can be very time-intensive, possibly requiring many hours or even days and thus eeectively making the database unavailable for unacceptable time spans. While researchers have looked at the deferred execution approach for schema evolution in an eeort to improve availability of the system, to the best of our knowledge ours is the rst eeort to provide a direct optimization strategy for a sequence of changes. In this paper, we propose heuristics for the iterative elimination and cancellation of schema evolution primitives as well as for the merging of database modiications of primitives such that they can be performed in one eecient transformation pass over the database. In addition we show the correctness of our optimization approach, thus guaranteeing that the initial input and the optimized output schema evolution sequence indeed produce the equivalent nal schema and data state. We also provide a proof of the algorithm's optimality by establishing the connuence property of our problem search space, i.e., we show that the iterative application of our heuristics always terminates and converges to a unique minimal sequence. To validate the feasibility of our optimization approach we have implemented our optimization strategy, the CHOP optimizer, on top of the Persistent Storage Engine (PSE), the Java-based object server developed by Object Design Inc. Moreover, we have conducted experimental studies that demonstrate the performance gains achieved by our proposed optimization technique over previous solutions. We would also like to thank our industrial sponsors, in particular, IBM for the IBM partnership award and Informix for software contribution. Special thanks also goes to the PSE Team speciically, Gordon Landis, Sam Haradhvala, Pat O'Brien and Breman Thuraising at Object Design Inc. for not only software contributions but also for providing us with a customized patch of the PSE Pro2.0 system that exposed schema-related APIs needed to develop our tool.