Energy-Efficient Task Scheduling for Heterogeneous Cloud Computing Systems

Abstract

Virtual machine migration and consolidation technologies can reduce the number of active physical machines (PMs) in the cloud platform, which can effectively reduce the power consumption of cloud platform. However, migration and consolidation of virtual machines (VMs) requires a suitable time to perform. When VMs' migration and consolidation are not performed, how to minimize power consumption of cloud platform through task scheduling is still one major challenge. Existing task scheduling methods mainly aimed at load balancing or minimizing task execution time, without considering the power consumption optimization of heterogeneous server cluster during task execution. To solve the high power consumption problem, this paper proposes a power-aware task scheduling (PATS) method for heterogeneous cloud platform. First, we construct a task scheduling model, which can aware the PM workload in real time according to the VM state, and predict the busy and idle power consumption of the PM, so as to obtain the power consumption of the cloud platform. This model also formulates the task scheduling optimization problem, which tries to minimize the power consumption of heterogeneous cloud platform. Second, we analyze the impact of PM type and task execution time on power consumption of cloud platform, and then propose a task scheduling algorithm to solve the above optimization problem. Finally, the algorithm is evaluated against other existing task scheduling algorithms under a CloudSim framework. Results demonstrate that PATS consumes 23.9-6.6% less total power consumption in comparison to the state-of-the-art.