摘要：在云数据中心网络中，海量的业务流在数据传输中往往会出现延迟、丢包等现象，但单用户性能测试无法实现多用户多业务流并发性的检测，为了能实时了解数据中心网络多用户多业务并发性数据的传输过程，发现数据传输过程中影响带宽、产生延迟、丢包现象的形成因素，解决了当前无法仿真多用户多条业务流测试瓶颈，解决不需增设服务器的情况下多用户并发性问题。文中采用Parallel Matrix Multiplication算法实现RDMA Ib_write源端口变化，解决了当前无法仿真多用户多条业务流测试瓶颈；采用基于RDMA技术的分布式和多线程组功能实现了并发测试，达到不增设服务器却能实现多用户并发的仿真效果，同时也降低了资源消耗及时间消耗。 In the cloud data center network, massive service flows often have delays and packet loss in data transmission, but single-user performance testing cannot achieve multi-user multi-service flow concurrency detection. The transmission process of multi-user multi-service concurrent data on the network has discovered the factors that affect the bandwidth, delay, and packet loss during data transmission, in the case of multi-user concurrency issues. In this paper, the Parallel Matrix Multi-plication algorithm is used to implement the RDMA Ib_write source port change, which solves the current test bottleneck of multi-user multi-service flow testing; the distributed and multi-threaded group function based on RDMA technology is used to achieve concurrent testing, and achieve the simulation effect of multi-user concurrency without adding a server, while also reducing resource consumption and time consumption.
关键词：云数据中心 ;RDMA ;多用户 ;多业务 ;并发性 ;Cloud Data Center ;RDMA ;Multi-User ;Multi-Service ;Concurrency
