集成SDN框架的启发式数据流调度算法研究  

HEURISTIC DATA FLOW SCHEDULING ALGORITHM INTEGRATING SDN FRAMEWORK

作　　者：黄润[1] 肖志良[1,2] Huang Run;Xiao Zhiliang(College of Electronic Information, Foshan Polytechnic, Foshan 528137,Guangdong, China;College of Information Management, Wuhan University, Wuhan 430072, Hubei, China)

机构地区：[1]佛山职业技术学院电子信息学院,广东佛山528137 [2]武汉大学信息管理学院,湖北武汉430072

出　　处：《计算机应用与软件》2019年第4期155-160,共6页Computer Applications and Software

基　　金：广东省自然科学基金项目(9451009001002711)

摘　　要：为了解决光数据中心的流调度问题和最大化云服务供应商的长期收入,提出最小拥塞和服务时间优先MC-STP(Minimum Congestion and Service Time Priority)的调度算法,以及基于拥塞的循环调度CBL(Congestion Based Loop)算法,并将其集成到软件定义网络(SDN)框架,以执行业务流调度和光路重构。其中:MC-STP向服务时间较短的业务流给予较高的优先级,使其先于其他流被容纳;CBL是为了弥补MC-STP的业务流饥饿问题,在计算出业务流的拥塞因子后,通过业务流的拥塞因子选择要调度的流,提供流之间的公平性。仿真结果表明,与端到端的调度算法、离散粒子群调度算法相比,该算法可明显降低拒绝率,提高波长利用率,有效提高云服务供应商的平均收入。In order to solve the flow scheduling problem of the optical data center and maximize the long-term revenue of cloud service providers, we proposed a minimum congestion and service time priority( MC-STP) scheduling algorithm and a congestion based loop scheduling( CBL) algorithm, which were integrated into a software defined network ( SDN) framework to perform traffic scheduling and optical path reconstruction. Among them, MC-STP gave higher priority to service flow with shorter service time, so that it could be accommodated before other flows. CBL was to make up for the starvation problem of MC-STP traffic. After calculating the congestion factor of traffic flow, we chose the flow to be scheduled through the congestion factor of traffic flow to provide fairness between flows. The simulation results verified the efficiency of the proposed algorithm. Compared with the end-to-end scheduling algorithm and discrete particle swarm optimization ( PSO), this algorithm can significantly reduce the rejection rate, improve the wavelength utilization, and effectively improve the average revenue of cloud service providers.

关 键 词：数据中心 调度算法 循环调度 最小拥塞 软件定义网络 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]