RHEV虚拟化平台网络流量分层性能优化实践

Layered performance optimization and practice of network traffic on RHEV virtualization platform

玉溪供电局RHEV虚拟化平台承载40余套业务系统,受原有网络环境的限制,平台不同逻辑网络共同使用千兆以太网链路,在运维中发现,当一台物理机宕机或物理节点需要维护时,虚拟机在线迁移速度缓慢,无法及时迁移到其它物理节点上。此外,在虚拟机迁移的过程中,由于需要占用大量网络带宽,不同逻辑网络之间存在流量争用,导致物理主机产生告警,甚至会出现迁移失败的情况,影响业务系统的可以性。为解决以上问题,本文提出将管理、虚拟机、显示和迁移四个逻辑网络进行分离,现了不同网络流量分层,解决网络带宽争用问题。同时将迁移网络升级为万兆网络,经验证,虚拟机迁移的效率得到了大幅提高,提高了平台的稳定性。

RHEV virtualization platform of Yuxi Power Supply Bureau carries more than 40 business systems. Due to the limitation of the original network environment, different logical networks of the platform use Gigabit Ethernet links together. During operation and maintenance, it is found that when a physical downtime or physical node needs to be maintained, the virtual machine migrates slowly online and cannot migrate to other physical nodes in time. In addition, in the process of virtual machine migration, due to the need to occupy a large amount of network bandwidth, there is traffic contention between different logical networks, which leads to alarms from physical hosts, even migration failure, affecting the viability of business systems. In order to solve the above problems, this paper proposes to separate the four logical networks of management, virtual machine, display and migration, realize different network traffic layers, and solve the problem of network bandwidth contention. At the same time, the migration network is upgraded to 10,000 Mbp network, which proves that the efficiency of virtual machine migration has been greatly improved and the stability of the platform has been improved.