Using Object - Oriented Principles to Optimize Update

View materialization is known to be a valuable technique for performance optimization in relational databases, and much work has been done addressing the problem of consistently maintaining relational views under update operations. However, little progress has been made thus far regarding the topic of view materialization in object-oriented databases (OODBs). In this paper, we demonstrate that there are several signiicant diierences between the relational and object-oriented paradigms that can be exploited when addressing the object-oriented view materialization problem. For example, we can use the subsumption relationships between classes to identify branches of classes to which we do not need to propagate updates. Similarly, we can use encapsulated interfaces combined with the fact that any unique database property is inherited from a single location to provide a \registration" service by which virtual classes can register their interest in speciic properties and be notiied upon modiication of those properties. We describe a number of techniques that take advantage of such data model diierences to optimize the maintenance of materialized views in the object-oriented model. We have successfully implemented all proposed techniques in the MultiView system, which provides updatable materialized classes and virtual schemata on top of the GemStone OODBMS. In this paper, we also report results from the experimental studies we have run on the MultiView system measuring the impact of various optimization strategies incorporated into our materialization update algorithms.