Network optimization method and system based on resource global affinity in numa virtualization environment

Abstract

The invention discloses an NUMA-based network optimization method and system for resource global affinity under a virtualized environment. The method comprises the three steps that 1, a hardware eventtriggered by the real-time access behavior of a virtual machine is monitored; 2, current virtual machine resource access affinity is calculated; 3, resources are dispatched according to the size of the affinity. The system comprises a virtual machine real-time behavior monitoring module, a host machine bottom layer hardware topological information obtaining module and a real-time virtual machineresource management module, the virtual machine real-time behavior monitoring module reflects the real-time access behavior of the virtual machine, the host machine bottom layer hardware topological information obtaining module provides the accurate dispatch basis for real-time dispatch, and the real-time virtual machine resource management module dispatches the virtual machine resources to the current optimal node to optimize the overall performance of the virtual machine. Accordingly, multicore resources and the performance of high-performance auxiliary acceleration hardware are efficientlyutilized, load of the system is effectively lowered, and the method and system adapt to application under the current network environment.

Description

technical field [0001] The present invention relates to computer system architecture, virtual machine resource scheduling, multi-core and virtual machine network performance optimization, and in particular to an optimization method and system for efficiently managing virtual machine resource scheduling to improve virtual machine network performance in a multi-core environment. Background technique [0002] Virtualization is a key technology in cloud computing. Virtualization technology can allow multiple operating systems to run on a physical server, and abstract the hardware resources (CPU, memory, I / O devices, etc.) of the physical machine into virtualized resources (virtual CPU, memory address mapping, virtual network card, etc. etc.), provided to customers for use. The use of virtualization technology has greatly reduced the investment of small enterprises in purchasing servers, and has also greatly improved the efficiency of idle hosts. Therefore, virtualization techno...

AI Technical Summary

Problems solved by technology

However, the NUMA architecture also brings another obvious disadvantage: there is a huge performance difference between local memory access and remote memory access. If the system-level application does not perceive the difference in NUMA data access, it will bring huge bandwidth performance. decrease, which affects the efficient operation of the application

[0006] 2. The existing virtual resource scheduler based on NUMA architecture does not take into account the characteristics of the underlying hardware topology

However, this assumption of equal NUMA node access is inconsistent with today's irregular application access patterns, so modeling with this assumption is inaccurate

Images