NNMonitor: Performance modeling for database servers

Abstract

Database Management Systems (DBMSs) are the cores of most information systems. Database administrators (DBAs) face increasingly more challenges due to the systems growing complexity and must be proficient in areas, such as capacity planning, physical database design, DBMS tuning and DBMS management. Furthermore, DBAs need to implement policies for effective workload scheduling, admission control, and resource provisioning. In response to these challenges we focus our attention on the development of online DBMS performance model. We aim to meet service level agreements (SLAs) and maintain peak performance for DBMS. To this end, we propose a neural network-based performance model called NNMonitor that can predict the performance metrics of DBMS online and determines if the DBMS needs to tune or not before entering into a complex tuning process. We make use of neural networks to build our proposed model taking into account the interaction among concurrently executing queries and predict throughput. The experimental evaluation demonstrates that this model is capable of predicting the performance metrics of real database servers with high accuracy.