An Adaptive Hybrid Controller for DBMS Performance Tuning

Abstract—Performance tuning process of database management system (DBMS) is an expensive, complex and time consuming process to be handled by human experts. A proposed adaptive controller is developed that utilizes a hybrid model from fuzzy logic and regression analysis to tune the memory-resident data structures of DBMS. The fuzzy logic module uses flexible rule matrix with adaption techniques to deal with fluctuations and abrupt changes in the operation environment. The regression module predicts fluctuations in operation environment so the controller can take former action. Experimental results on standard benchmarks showed significant performance enhancement as compared to built-in self-tuning features. Keywords—automatic database tuning; fuzzy logic; adaptive controller; regression; self-tuning; DBMS I