论文部分内容阅读
当前数据中心网络数据流量大小分布不均衡,传统等价多路径转发ECMP算法容易将多条大数据流转发至同一链路,导致链路瓶颈.提出面向CLOS结构数据中心网络的基于SDN的流量分类路由SCR(SDN-Based Classified Routing)机制.SCR利用Open Flow机制周期性统计数据流来计算阈值,通过动态判定方法将数据流分为大流和小流.大流由SDN控制器通过自适应路由算法计算路径,小流由交换机通过流量无视路由算法计算路径.本文在VL2架构上建立合成流量模型对SCR进行性能分析与评价.仿真与分析结果表明,与ECMP相比,SCR在网络吞吐量、数据流丢弃率、分组端到端时延和平均队列长度等方面具有优势.
Currently, data traffic in data center networks is unevenly distributed, and the ECMP algorithm is easy to forward multiple large data flows to the same link, resulting in link bottlenecks.This paper proposes SDN-based traffic classification for CLOS data center networks Based on SDN-Based Classified Routing (SCR) mechanism, SCR uses Open Flow mechanism to calculate periodic thresholds and divides the data flow into high and low flows by dynamic decision.A large flow is controlled by SDN controller through adaptive routing Algorithm to calculate the path, the flow of small flow by the switch ignoring the routing algorithm to calculate the path by traffic.This paper establishes a synthetic flow model on the VL2 architecture for SCR performance analysis and evaluation.The simulation and analysis results show that compared with ECMP, SCR in the network throughput, Data flow discard rate, packet end-to-end delay and average queue length and other advantages.