The architecture of BrAID: a system for bridging AI/DB systems

The design of BrAID (a bridge between artificial intelligence and database management systems), an experimental system for the efficient integration of logic-based artificial intelligence (AI) and databases (DB) technologies, is described. Features provided by BrAID include (a) access to conventional DBMSs, (b) support for multiple inferencing strategies, (c) a powerful caching subsystem that manages views and uses subsumption to facilitate the reuse of previously cached data, (d) lazy or eager evaluation of queries submitted by the AI system, and (e) the generation of advice by the AI system to aid in cache management and query execution planning. Some of the key aspects of the BrAID architecture are discussed, focusing on the generation of advice by the AI system and its use by a cache management system to increase efficiency in accessing remote DBMSs through the selective application of such techniques as prefetching, query generalization, result caching, attribute indexing, and lazy evaluation.<<ETX>>