服务网格性能优化关键技术研究

RESEARCH ON PERFORMANCE OPTIMIZATION KEY TECHNOLOGY FOR SERVICE MESH

服务网格(Service Mesh)是用于处理服务间通信的专用基础设施层。为了有效提升其性能,从并发处理模型和交互协议两个方面研究服务网格的性能优化方法,分别提出基于Reactor设计模式的服务网格高并发处理模型和面向服务网格的轻量级、自定义的代理交互协议,并实现一个服务网格原型系统。通过实验与当前主流的开源服务网格系统Linkerd和Envoy进行性能对比分析,实验结果表明该方法能够显著提升服务网格的并发处理性能,使实现的原型系统在并发处理和通信方面都有良好的性能表现。

The service mesh is a dedicated infrastructure layer for handling communication between services. To effectively improve the service mesh performance, this paper presents an approach to optimizing service mesh performance from two aspects. The approach proposed a high-concurrency processing model based on the Reactor design pattern and a customized light-weight protocol for communications between service mesh agents respectively. This approach implemented a prototype and conducted experiments to evaluate the approach’s effectiveness by comparing the prototype with some popular open-source service mesh systems like Linkerd and Envoy. The experimental results show that this approach effectively improves the overall performance of a service mesh in terms of query per second and communication latency.