Optimization Of Resource Affinity Based On KVM In High Performance Network

With the continuous improvement of hardware network,especially in the cloud computing environment,the affinity of resources under virtualization has become an important and urgent research topic.The concept of affinity optimization refers to the "affinity relationship" in the physical architecture to improve the resource utilization.This paper focus on KVM virtual environment and make full use of SR-IOV network card virtualization technology to analyze the Non Uniform I/O Access architecture,and the key factors of resource affinity under high-performance network,and explore our research.This paper presents a set of named vNIO containing the virtual machine placement and the virtual machine dynamic scheduling system,which can achieve virtual machine based on different physical resource usage,optimize the placement strategy and scheduling algorithm,and achieve a better affinity performance including VCPU,Virtual memory and Virtual I / O.The performance has large difference under different placement strategies by testing our placement algorithms,especially in the high-performance network environment.As the network card speed rising,the differences in resources is also growing.Which proves the effectiveness of the research,that is,in high-performance network with virtual environment,it is necessary to fully consider the CPU,memory,I / O devices in order to enhance the system affinity.The scheduling algorithm has a certain threshold to enhance the affinity of physical resources.The system performance improved by up to 20% with a small number of virtual machines.With up to 16 VMs,the algorithm can increase performance by about 5%.Overall,the research can effectively improve overall performance.