基于SDN的数据中心流量负载均衡  

DATA CENTER TRAFFIC LOAD BALANCING BASED ON SDN

作　　者：杨耀忠 刘宝军 段鸿杰 罗阳 程鹏[2] Yang Yaozhong;Liu Baojun;Duan Hongjie;Luo Yang;Cheng Peng(Information Management Center of Sinopec Shengli Oilfield,Dongying 257000,Shandong,China;College of Computer Science and Technology,China University of Petroleum(East China),Qingdao 266580,Shandong,China)

机构地区：[1]胜利油田分公司信息化管理中心,山东东营257000 [2]中国石油大学(华东)计算机科学与技术学院,山东青岛266580

出　　处：《计算机应用与软件》2022年第11期130-135,147,共7页Computer Applications and Software

摘　　要：随着云计算技术和分布式服务的发展,数据中心的“东西向”大象流量急剧增加。以往的研究表明,如何有效地调度大象流量成为数据中心保持高性能和避免网络拥塞的重要问题。针对上述问题,研究了数据中心的流量调度问题,重点是大象流量调度。基于软件定义网络(SDN)和稳定匹配理论对数据中心大象流量调度进行建模,提出流量调度算法TPLM(Traffic Priority Level Matching),TPLM依据流数据速率和交换机内存容量之间的关系得到最优调度。仿真结果表明,TPLM减少了平均流完成时间,提高了系统吞吐量。With the development of cloud computing technology and distributed services,the"east-west"elephant traffic in the data center has increased dramatically.Previous studies have shown that how to effectively arrange elephant traffic becomes an important issue for data centers to maintain high performance and avoid network congestion.In response to the above problems,we studied the traffic scheduling problem in the data center,especially the elephant traffic scheduling.Then the data center elephant traffic scheduling based on software-defined network(SDN)and stable matching theory was modeled,and the traffic scheduling algorithm TPLM(traffic priority level matching)was proposed.TPLM obtained the optimal scheduling based on the relationship between flow data rate and switch memory capacity.The simulation results show that TPLM reduces the flow average completion time and improves the system throughput.

关 键 词：流量调度 软件定义网络 数据中心网络 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]