【摘要】 为了满足云用户请求响应要求的同时进一步提高云平台能量效率,融合唤醒阈值与半休眠模式,提出一种新型的云虚拟机调度策略。在一个半休眠周期结束时刻,如果缓冲区中等待的云用户请求数达到阈值N,虚拟机则由半休眠状态转入唤醒状态,以正常速率为云用户请求提供服务;否则,虚拟机将重新开始一次新的半休眠周期,持续为云用户请求提供低速服务。根据云虚拟机调度策略的原理,建立一个具有N策略和多重异步工作休假的多服务台排队模型。利用拟生灭过程和矩阵几何解方法,推导云用户请求平均逗留时间与系统节能率等性能指标。综合数值分析实验和系统仿真实验,评估云虚拟机调度策略的系统性能。构造成本函数,利用蚁群智能寻优算法,给出云虚拟机调度策略的优化方案。更多还原

【Abstract】 In order to meet the response performance of cloud requests,as well as further improve the energy efficiency of the cloud platform,a novel virtual machine(VM) scheduling strategy with awakening threshold and semi-sleep mode is proposed.At the end instant of a semi-sleep period,if the number of cloud requests waiting in the buffer reaches the threshold N,the VM will convert to awake state from semi-sleep state,and then provide service at the normal service rate.Otherwise,the VM will restart a new semi-sleep period and continue to provide service at the lower service rate.According to the principle of the proposed strategy,a multi-server queueing model with N-policy and multiple asynchronous working vacations is established.Using the quasi-birth-and-death processes and the matrix-geometric solution method,the performance measures,such as the average latency of cloud requests and the energy-saving rate of system,are derived.Experimental results with analysis and simulation are provided to evaluate the system performance of the proposed strategy.In addition,by constructing the system cost function and using ant colony intelligent optimization algorithm,the optimization scheme of the VM scheduling strategy is presented.更多还原