摘要
软件定义网络(software defined network,SDN)分布式控制平面的负载均衡主要通过交换机迁移实现,但现有算法仅考虑了CPU一维控制资源,且基于集中式算法实现,复杂度较高。针对这一问题,将交换机迁移决策抽象为在三维资源(CPU、带宽和存储)限制下的资源效用最大化问题;然后将其等效为零和博弈中的收益最大化问题,并设计了一种合成分布式博弈算法;最后在Beacon控制器实现了该算法。实验结果显示该方案的流表安装时间具有最低振荡曲线,且时间都小于100 ms,近似最优方案。
The load balancing of distributed control plane in SDN ( software defined network) is mainly conducted through switch migration. However, existing switch migration algorithms only considered the CPU resource and were implemented in centralized style which was complicated. This paper modeled the switch migration problem as a resource utility maximization problem under the restrictions of the three dimensional resources ( CPU, bandwidth and storage), and then equivalented it to revenue maximization problem in a zero-sum game. This paper also designed a distributed solution called synthesizing distributed game algorithm and implemented it in the Beacon controller. Evaluation results show that flow setup time has the lowest oscillation curve,which is less than 100 ms and approximates the optimal solution.
出处
《计算机应用研究》
CSCD
北大核心
2017年第10期3045-3050,3056,共7页
Application Research of Computers
基金
国家"973"计划资助项目(2012CB315901
2013CB329104)
国家自然科学基金资助项目(61309020
61372121)
国家"863"计划资助项目(2015AA016102)
关键词
软件定义网络
交换机迁移
非合作博弈
弹性控制
可扩展性
software defined network
switch migration
non-cooperative game
elastic control
scalability