面向高并发的消息中间件设计与实现

Design and Implementation of Message Oriented Middleware for High Concurrency

针对高并发场景中突发流量过大宕机等典型问题,提出一种面向高并发场景的消息中间件进行流量消峰。该消息中间件采用Producer-Broker-Consumer架构,划分为通用层、网络通信层、消息服务器、消息客户端;自定义通信协议,采用Push模式、负载均衡轮询方式进行消息推送等。将该中间件与主流消息中间件对比分析并应用到商品秒杀平台中,运用JMeter工具进行压力测试,验证流量消峰综合性能。实验结果表明,提出的面向高并发场景的消息中间件相比主流消息中间件具有较好的流量消峰性能;该消息中间件应用到商品秒杀平台中可以有效地实现流量消峰,当错误请求率为0的情况下并发量达55 802/sec,吞吐量达6 319.8/sec不宕机。

Regarding the typical problems such as downtime caused by excessive sudden traffic in high concurrency scenarios, a message oriented middleware was proposed to eliminate traffic peaks. The message middleware adopts ProducerBroker-Consumer architecture, which was divided into general layer, network communication layer, message server and message client. It defined the communication protocol, used push mode and loaded balancing polling mode to push messages. This paper compared it with the mainstream message oriented middleware and applied it to the commodity spike platform, then used JMeter to conduct pressure test to verify the comprehensive performance of traffic peak elimination. The experimental results showed that the middleware had better traffic peak elimination performance than the mainstream message middleware. By applying it to the commodity spike platform, the message middleware can effectively eliminate the peak traffic. When the error request rate was 0, the concurrency reached 55 802/sec and the throughput reached 6 319.8c without downtime.