云数据中心基于负载预测的物理主机状态检测策略

针对云数据中心现有物理主机状态检测算法对提高云数据中心物理资源的利用率效果不明显问题,提出了基于负载预测的物理主机状态检测策略(load prediction based physical host status detection,LP-PHSD),LP-PHSD利用时间序列和二次指数平滑法预测出物理主机在未来一段时间内的资源利用率情况,同时结合绝对中位差方法,确定资源利用率动态阈值边界,选择适当的时刻进行迁移,提高物理资源的利用效率,降低能量消耗.LP-PHSD包括源物理主机状态检测和目标物理主机状态检测2个部分,可以很好地判断出虚拟机迁移的时刻.实验表明,经LP-PHSD策略优化后的新虚拟机迁移方法与近几年的BenchMark迁移模型比较起来,云数据中心的总体能量消耗降低,虚拟机迁移次数减少,云服务质量明显提高.

Aload prediction based physical host status detection strategy in cloud data centers

Current physical host status detection algorithm cannot easily promote the utilization efficiency of physical resource in cloud data centers. A load prediction based physical host status detection strategy called LP-PHSD is proposed in this paper. LP-PHSD uses time-series and double exponential smoothing forecasting technique. Combined with Median Absolute Deviation method, the resource utilization thresholds in physical host which make LP-PHSD is dynamic and adaptive.LP-PHSD is an algorithm which can decide whether it is really necessary to migrate a VM depending on the present as well as future load and thus the resource utilization efficiency can be increased and energy consumption has been saved.LP-PHSD can exclude two kinds of physical hosts, source to be migrated physical host and destination to be migrated physical hosts and thus the accurate decision of VM migration has been obtained. Experimental results show that compared with benchmark VM migration models, energy consumption and numbers of VM migration has been reduced in LP-PHSD. The Qos of cloud service provider has been also promoted.