An improved artificial bee colony algorithm to minimum propagation latency and balanced load for controller placement in Software Defined Network
Yurui Ma,Jue Chen,Wenjing Lv,Xihe Qiu,Yue Zhang,Wanxiao Liu
DOI: https://doi.org/10.1016/j.comnet.2024.110600
IF: 5.493
2024-06-15
Computer Networks
Abstract:Software Defined Network (SDN) has become an increasingly popular network architecture as it provides flexible global network management by decoupling the control plane from the data plane, and deploys multiple controllers in a logically centralized and physically distributed manner in the network. In this architecture, controllers serve as the brains and are responsible for directing traffic and optimizing performance. However, deploying SDN controllers can be a complex process that requires careful planning and consideration, resulting in the emergence of the Controller Placement Problem (CPP). Considering that this problem is proven to be NP-hard, this paper proposes an algorithm named improved Artificial Bee Colony with K-means (ABCK) algorithm to calculate the optimal or near-optimal solution with high efficiency. On the basis of the heuristic algorithm, i.e., Artificial Bee Colony (ABC) algorithm, we adjust the initial food source search equation by using approach of K-means, and optimize related parameters to minimize the propagation latency between controllers and switches (including average latency and worst case latency) as well as maintaining load balancing among controllers. In order to prove the effectiveness of our proposed method, we conduct experiments on five real topologies from the Internet Topology Zoo and Internet2 OS3E to compare the performance between ABCK and other state-of-the-art methods including ABC, cross entropy, K-means, K-means++, Pareto Simulated Annealing (PSA) and Pareto Bacterial Foraging Optimization (ABFO). The results verify that our proposed approach can realize the minimum propagation latency (including the average and worst case latencies) for different network scales with different numbers of controllers among all the methods, with no more than 4.37 % margin from the optimal solution. At the same time, by using our approach, the load of controllers is the most balanced and the computation time is the least among all the methods. In addition, the simulation results show that our proposed method performs better in large scale network topologies.
computer science, information systems,telecommunications,engineering, electrical & electronic, hardware & architecture