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
李梦桃 1徐秀丽1
作者信息
- 1. 燕山大学 理学院,河北 秦皇岛 066004
- 折叠
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.
Abstract
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 num-ber 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 ma-trix-geometric solution method,and a reasonable benefit function is constructed to discuss the profit problem of the cloud system.Fi-nally,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Key words
cloud computing/wake-up mechanism/Min(c,V)policy/active threshold optimization引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据