对延迟敏感型应用的虚拟机调度算法

摘要

This paper proposes two improvements to the problem that the Credit scheduling algorithm can not guarantee the real-time.First,we can shorten the system respond time by loading balancing about the virtual CPU in BOOST state when it has a lot of I/O tasks.Second,we can use dynamic time slice instead of the former fixed time slice to adapt to the dynamic change of the virtual CPU.The impact of the improvement on the I/O task is evaluated by detecting the average response time and turnaround time of the task.The experimental results show that the average response time of the improved Credit scheduling algorithm is 102.3％ lower than that before improvement,which can significantly enhance the performance of I/O latency-sensitive applications.%针对Credit调度算法不能保证实时性的不足提出两点改进.首先,当有大量I/O任务时对BOOST态虚拟CPU进行负载均衡来缩短系统响应时间.其次,利用动态时间片代替原来的固定时间片去适应虚拟CPU的动态变化.通过检测系统对任务的平均响应时间和周转时间来评估改进前和改进后对I/O任务的响应的影响.实验研究表明,改进之后的Credit调度算法平均响应时间与改进前相比降低了102.3％,可以显著提高I/O延迟敏感型应用的性能.