Phase-Based Performance Optimization Utilizing the Dynamic Behavioral Changes in Applications

Abstract

As the big data and application complexities increase, the dynamic performance optimization is becoming a very significant issue. Auto-tuning of the execution of applications has gained a lot of attention recently. However, the exploitation of the dynamic behavioral changes in performance optimization still represents a challenge. This paper proposes a phase-based optimization approach by utilizing the dynamic behavioral changes in applications. This approach aims at enabling each application’s phase to adapt its resource usage based on the dynamic needs of the resources during its execution. Through a priori and posteriori knowledge of the application, operating systems could expose the characteristics of the application’s behavior and the actual resource requirements of each phase. The results of the comparative evaluation with a previous work show, significant improvements in performance with less resource usage, and more convergent results which guarantees the quality of service.