智慧协同网络中基于流量矩阵的负载均衡路由机制  

Traffic matrix-based load balancing routing in flow-based smart identifier network

作　　者：贾濡 郜帅[1] 罗洪斌[1] 张宏科[1] 万明[2] 

机构地区：[1]北京交通大学电子信息工程学院下一代互联网互联设备国家工程实验室,北京100044 [2]中国科学院沈阳自动化研究所网络化控制系统重点实验室,辽宁沈阳110016

出　　处：《通信学报》2016年第4期128-138,共11页Journal on Communications

基　　金：国家重点基础研究发展计划("973"计划)基金资助项目(No.2013CB329100);国家自然科学基金资助项目(No.61232017;No.61271200;No.61501447);国家高技术研究发展计划("863"计划)基金资助项目(No.2015AA011906)~~

摘　　要：智慧协同网络具有能够实时准确测算流量矩阵的特点。将流量矩阵作为约束,对负载均衡路由优化问题进行建模,利用拉格朗日对偶方法,将原问题转化为优化目标易实现的对偶问题。为实现对偶问题优化目标,提出一种基于流量矩阵的负载均衡路由(TM-LB,traffic matrix based load balancing)算法,供控制层根据实时网络情况为后续流规划传输路径。利用OMNET++仿真器在NFSnet拓扑结构上进行仿真实验,结果表明TM-LB相比传统路径规划机制能有效避免拥塞,实现负载均衡。最后,搭建原型系统对TM-LB算法的开销进行测试。Smart identifier network（SINET） has an advantage of being aware of the accurate traffic matrix. Traffic matrix was use as the constraint, model on load balancing routing problem, and transform the primal problem into dual problem by using Lagrange function. Therefore, the optimization goal of the dual problem can be easily achieved. In order to achieve the dual problem optimization goal, a traffic matrix based load balancing routing algorithm（TM-LB） was proposed. By performing the TM-LB, the control plane can plan paths for subsequent traffic according to network fluctuation. OMNET＋＋ was used to run the experiment simulation based on NSFnet topology. The results show that TM-LB can better reduce congestion and realize load balancing, compared with traditional routing mechanism. Finally, a proof-of-concept was built implementation and carry out experiments for testing the overhead of TM-LB algorithm.

关 键 词：智慧协同网络 流量矩阵 路由优化 拉格朗日对偶 负载均衡 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]