Collaborative Agents Supported Automatic Physical Database Design Based on Description Logics Reasoning

Physical database design is a key phase in devising and deploying a database system to improve overall system performance, and automatic physical database design becomes an important value-added feature in a database system to decrease the TCO (total cost of ownership). A GAPDD (general automatic physical database design) framework, with two phases, the workload mining phase and the design advising phase, is provided to support automatic physical database design for relational database systems. Three collaborative agents, the workload mining agent (WMA), the feature generation agent (FGA), and the optimal search agent (OSA) work together in GAPDD. The feature generation procedure in FGA is further studied, and a reasoner based on description logics (DL) is offered to reason useful features. An index advisor for OSCAR relational engine is prototyped, and experiment results show its feasibility and efficiency