Virtual network performance optimization method based on OpenStack control plane

Abstract

The invention belongs to the field of virtual network performance in a cloud network, and relates to a virtual network performance optimization method based on an OpenStack control plane. The method comprises the following steps: improving a sending mechanism of the RPC from the perspective of a message generator, and only sending the RPC message to an L2 agent under the same subnet, so that the sending quantity of the RPC is reduced; and adopting a hybrid message queue mode from the perspective of a message processor. The persistence requirements of notification messages can be ensured by utilizing the characteristics of RabbitMQ persistence and high availability, and a large number of RPC messages can be rapidly processed in a direct communication mode by utilizing the characteristics ofZeroMQ rapidness and low delay, so that RPC message accumulation is reduced; and, on the premise of not influencing the original network function, the RPC message is independent and is processed by ZeroMQ, intermediate agency is not needed, the release is quicker, the tenant network communication performance is improved from two aspects, and the performance of the virtual network is further improved.

Description

technical field [0001] The invention relates to the field of virtual network performance in cloud networks, in particular to a method for optimizing virtual network performance based on an OpenStack control plane. Background technique [0002] In recent years, driven by technological development, market demand, and business model transformation, cloud computing has developed rapidly. It integrates scattered network, computing, and storage resources, and combines virtualization technology to create a virtual network for tenants in the form of resource leasing. Provide virtual resources and services. Compared with the traditional network environment that relies on physical servers, the cloud environment has lower cost, high scalability, convenient and efficient management, high resource utilization, and a complete migration disaster recovery backup solution, which can realize intelligent resource scheduling and unified and flexible management. Manufacturers can also implement...

AI Technical Summary

Problems solved by technology

[0008] The port status change of the virtual machine is scanned by the L2 proxy and reports the update_port_status information through RabbitMQ. The L2pop driver fans out the fdb_entries message to the L2 proxy of all computing nodes. When the network scale is slightly large, RabbitMQ needs to carry and process a large number of fan-out messages ; Too many RPC messages will cause message accumulation. In severe cases, the RPC message storm caused by too many RPC messages will even cause the message queue to crash. The message queue becomes the performance bottleneck of virtual network communication, which seriously affects the performance and scalability of the OpenStack environment. ; After the L2 population mechanism is enabled for Layer 2 communication, although the broadcast storm is alleviated, the release of RPC messages depends on the message queue, which brings huge pressure to the message queue; when the network scale increases, the message queue becomes the performance bottleneck of the virtual network. Seriously impact network performance and scalability

Images