CloudFreq: Elastic Energy-Efficient Bag-of-Tasks Scheduling in DVFS-Enabled Clouds

Energy consumption imposes a significant cost for data centers in providing cloud services. Many studies explore the opportunities to save power by energy-efficient task scheduling based on the technique of dynamic voltage and frequency scaling (DVFS). However, most of them assume that energy budgets and/or deadline constraints are known in advance. But these information can hardly be acquired in general computing environments, such as cloud computing, and job rejections caused by restricted constraints are intolerable to guarantee the service-level agreement (SLA). Moreover, previous works prefer to provide “black-box” algorithms with little consideration on adjustability, and cannot satisfy runtime requirements in performance and energy-saving. This paper proposes an elastic energy-efficient algorithm called CloudFreq for bag-of-tasks scheduling in DVFS-enabled clouds. CloudFreq enables a model of elastic, adjustable energy-efficient scheduling without any prior knowledge of constraints, and then eliminates job rejections accordingly. CloudFreq also provides an entry for operators to scale system performance at runtime. Experimental results demonstrate that the proposed algorithm can effectively perform energy-efficient scheduling without constraints, and has the capability of making an appropriate tradeoff to improve the weighted balance between schedule length and energy-saving.