基于Min(c,V)唤醒机制和活跃阈值的云虚拟机调度优化  

Scheduling Optimization of Cloud Virtual Machines Based on Min(c,V)Wake-up Mechanism and Active Threshold

作　　者：李梦桃 徐秀丽[1] LI Mengtao;XU Xiuli(School of Science,Yanshan University,Qinhuangdao 066004,China)

机构地区：[1]燕山大学理学院,河北秦皇岛066004

出　　处：《山西大学学报（自然科学版）》2024年第5期943-953,共11页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金项目(62171143)。

摘　　要：为了满足云用户对系统响应性能的需求,同时进一步提高云系统的节能水平,论文提出一种新型虚拟机调度策略。当虚拟机处于休眠状态时,一旦缓冲区内的云任务数超过服务台数c或者休眠定时器到时,所有的虚拟机便立即停止休眠进入唤醒状态。在唤醒期结束时,如果缓冲区内等待的云任务数达到阈值N,虚拟机由唤醒状态转入活跃状态,否则转入空闲状态等待云任务到达。基于上述背景,构建具有Min(c,V)策略与活跃阈值N的可变到达速率的多服务台排队模型。利用矩阵几何解方法,推导出云系统稳态性能指标,并构建合理的收益函数讨论云系统收益问题,综合数值分析验证了所提策略的有效性。In order to meet the demand of cloud users for system response performance and further improve the energy saving level of cloud system,this paper proposes a new virtual machine(VM)scheduling policy.When VMs are in the sleep state,once the number of cloud tasks in the buffer exceeds the number of service stations c or the sleep timer ends,all VMs immediately stop sleeping and enter the wakeup state.Moreover,at the end of the wake-up period,if the number of waiting cloud tasks in the buffer reaches the threshold N,the VMs are transferred from the wake-up state to the active state.Otherwise,they are transferred to the idle state to wait for cloud tasks to arrive.Based on the above background,a multi-server queuing model with variable arrival rate and Min(c,V)mechanism and active threshold N is constructed.The steady-state performance indices of the cloud system are derived by using matrix-geometric solution method,and a reasonable benefit function is constructed to discuss the profit problem of the cloud system.Finally,the effectiveness of the proposed policy is verified by comprehensive numerical analysis.

关 键 词：云计算 唤醒机制 Min(c V)策略 活跃阈值优化 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]