EPOBF: Energy Efficient Allocation of Virtual Machines in High Performance Computing Cloud

Abstract

Cloud computing has become more popular in provision of computing resources under virtual machine (VM) abstraction for high performance computing (HPC) users. A HPC cloud is such a cloud computing environment. One of the challenges of energy-efficient resource allocation of VMs in HPC clouds is the trade-off between minimizing total energy consumption of physical machines (PMs) and satisfying Quality of Service (e.g. performance). On the one hand, cloud providers want to maximize their profit by reducing the power cost (e.g. using the smallest number of running PMs). On the other hand, cloud customers (users) want highest performance for their applications. In this paper, we study energy-efficient allocation of VMs that focuses on scenarios where users request short-term resources at fixed start-times and non-interrupted durations. We then propose a new allocation heuristic (namely Energy-aware and Performance-per-watt oriented Best-fit (EPOBF)) that uses performance-per-watt as a metric to choose which most energy-efficient PM for mapping each VM (e.g. the maximum of MIPS/Watt). Using information from Feitelsons Parallel Workload Archive to model HPC jobs, we compare the proposed EPOBF to state-of-the-art heuristics on heterogeneous PMs (each PM has multicore CPUs). Simulations show that the proposed EPOBF can significantly reduce total energy consumption when compared with state-of-the-art allocation heuristics.