Sustaining software interoperability via shared, evolving object repositories: system optimization and evaluation

Powerful interoperability-enabling solutions for software application integration must allow applications to evolve and data requirements to change, while minimizing such changes on other integrated applications. Thus, we have developed the transparent schema evolution (TSE) system that accomplishes evolution by generating a new object-oriented view schema to capture the changes desired by the user, while preserving existing view schemas for old applications. This generation of a potentially large number of schema versions over time results in an excessive build-up of classes and underlying object instances - some of which may no longer be in use. The presence of no-longer-utilized view schemas will result in storage overhead costs and a degradation of system performance due to the update propagation of our materialized view foundation. In this paper, we address this problem using consistent schema removal techniques. Our solution is based on a formal model of capturing all dependencies between classes as logic clauses and of manipulating them to make decisions on class deletions and nondeletions while guaranteeing the consistency of the schema. Based on this formal model, we have developed and proven consistent a dependency graph (DG) representation and associated set of rules for DG generation, reduction, and transformation. The later has been successfully implemented in our TSE optimizer tool. To address the problem that view schema removal is sensitive to the order in which individual classes are processed, we present a cost model for evaluating alternative removal patterns on DG. We also report our preliminary experimental studies that validate our approach and demonstrate its impact on the performance of the TSE system.