Chaining, Referral, Subscription, Leasing: New Mechanisms in Distributed Query Optimization

This work outlines a exible framework for optimizing and deploying distributed queries in wide area networks. The database eld has developed very powerful techniques for nding eÆcient execution plans for declaratively specied queries. However, applying these optimization techniques in the setting of distributed information management requires centralized knowledge of the entire network and assumes passive behavior from the data sources. The reality of the Web is dierent. Future distributed query optimizers must handle (in fact, exploit!) a rich variety of information ow mechanisms like chaining, referral, proxying, brokering, publish-subscribe, leasing, etc. We look to mobile agent technologies for the combination of exibility and precision needed for handling these mechanisms. Our language-based approach uses a mobile process calculus based on the pi-calculus in combination with a powerful query-plan language. The salient characteristic of the language is that messaging, migration, and database operations all live in the same semantic space and interact, creating new opportunities for optimization.