Metaheuristic Solutions for Solving Controller Placement Problem in SDN-based WAN Architecture
Metaheuristic Solutions for Solving Controller Placement Problem in SDN-based WAN Architecture
复制标题
DOI：
10.5220/0006483200150023
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
Kshira Sagar Sahoo;Anamay Sarkar;S. Mishra;B. Sahoo;Deepak Puthal;M. Obaidat;B. Sadoun
中科院分区：
文献类型：
--
作者：
Kshira Sagar Sahoo;Anamay Sarkar;S. Mishra;B. Sahoo;Deepak Puthal;M. Obaidat;B. Sadoun
软件定义网络（SDN）是现代网络系统中的一种流行范例，它将控制逻辑与底层硬件设备解耦。控制逻辑已实现为软件组件，并驻留在称为控制器的服务器中。为了提高性能，在大规模网络中部署多个控制器是SDN的关键挑战之一。为了解决这个问题，作者将控制器放置问题（CPP）视为多目标组合优化问题，并使用不同的启发式方法。对于中小型拓扑，此类启发式方法可以在特定时间范围内执行，但对于广域网 (WAN) 等大规模实例，则超出了范围。为了获得更好的结果，我们提出了粒子群优化（PSO）和萤火虫两种基于群体的元启发式算法来实现控制器的最佳放置，它们采用一组特定的目标函数并从中返回最佳可能位置。该问题已被定义，同时考虑控制器切换和控制器间延迟作为目标函数。在一组公开可用的网络拓扑上根据执行时间评估算法的性能。结果表明，FireFly 算法在各种条件下均优于 PSO 和随机方法。
Software Defined Networks (SDN) is a popular paradigm in the modern networking systems that decouples the control logic from the underlying hardware devices. The control logic has implemented as a software component and residing in a server called controller. To increase the performance, deploying multiple controllers in a largescale network is one of the key challenges of SDN. To solve this, authors have considered controller placement problem (CPP) as a multi-objective combinatorial optimization problem and used different heuristics. Such heuristics can be executed within a specific time-frame for small and medium sized topology, but out of scope for large scale instances like Wide Area Network (WAN). In order to obtain better results, we propose Particle Swarm Optimization (PSO) and Firefly two population-based meta-heuristic algorithms for optimal placement of the controllers, which take a particular set of objective functions and return the best possible position out of them. The problem has been defined, taking into consideration both controllers to switch and inter-controller latency as the objective functions. The performance of the algorithms evaluated on a set of publicly available network topologies in terms execution time. The results show that the FireFly algorithm performs better than PSO and random approach under various conditions.