基于任务分类的虚拟CPU调度模型

为了桥接语义鸿沟，提升I/O性能，需要对执行不同类型负载的虚拟CPU（vCPU）采取不同的调度策略，故而虚拟CPU调度算法亟需优化。基于KVM虚拟化平台提出一种基于任务分类的虚拟CPU调度模型STC（virtual CPU scheduler based on task classification），它将虚拟CPU（vCPU）和物理CPU分别分为两个类型，分别为short vCPU和long vCPU，以及short CPU 和long CPU，不同类型的vCPU分配至对应类型的物理CPU上执行。同时，基于机器学习理论，STC构建分类器，通过提取任务行为特征将任务分为两类，I/O密集型的任务分配至short vCPU上，而计算密集型任务则分配至long vCPU上。STC在保证计算性能的基础上，提高了I/O的响应速度。实验结果表明，STC与系统默认的CFS相比，网络延时降低18%，网络吞吐率提高17%~25%，并且保证了整个系统的资源共享公平性。

Scheduling model of virtual CPU based on task classification

In order to bridge the semantic gap and improve the performance, different scheduling strategies should be applied to virtual CPUs(vCPU) which execute different types of tasks. Thus, the scheduling of vCPU should be optimized. This paper ploposed the STC(virtual CPU scheduler based on task classification), a scheduling model of virtual CPUs which was based on task classification. In STC, vCPUs and physical CPUs were classified into two types, that were short vCPU and long vCPU, which were accordingly mapped to short CPU and long CPU. Moreover, STC built classifier based on machine learning, and tasks were classified into I/O-bound ones and CPU-bound ones, which were allocated to short vCPUs and long vCPUs. STC improved the I/O responding speed without influencing the computing performance. Compared with default CFS algorithm, the experiment results show that STC has achieved time delay 18% decrease, bandwidth 17%~25% improvement, and ensures the fairness of the whole system.