Identifying New Directions in Database Performance Tuning

Abstract

Database performance tuning is a complex and varied active research topic. With enterprise relational database management systems still reliant on the set-based relational concepts that defined early data management products, the disparity between the object-oriented application development model and the object-relational database model, called the object-relational impedance mismatch problem, is addressed by techniques such as object-relational mapping (ORM). This, compounded with changes in the way data is produced, stored and managed can result in generally poor query performance for SQL produced by object-oriented applications and an irregular fit with cost-based optimisation algorithms, and leads to questions about the need for the relational model to better adapt to a more diverse set of queries. This paper discusses existing database performance optimisation techniques and approaches and makes the argument that current database performance tuning approaches need revisiting to support queries developed through ORM tools. This paper also introduces our current research, which includes exploring concepts such as dynamic schema redefinition; query analysis and optimisation modelling driven by machine learning; and augmentation of the cost-based optimiser model.