Controller placements for latency minimization of both primary and backup paths in SDNs

• Published in: Computer communications Vol. 163; pp. 35 - 50
• Main Authors: Fan, Yuqi, Wang, Lunfei, Yuan, Xiaohui
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.11.2020
• Subjects: Controller placement; Latency; Multi-objective optimization; Reliability; SDN; Single link failure; Single link failure; Multi-objective optimization; SDN; Reliability; Controller placement; Latency
• ISSN: 0140-3664; 1873-703X
• DOI: 10.1016/j.comcom.2020.09.001

Software-defined networking (SDN) is a revolutionary network architecture that separates the network control layer from the underlying equipment. Multiple controllers form a logically centralized control layer in large-scale networks, which raises the controller placement problem. Most of the research on latency-oriented controller placement optimized the delay between switches and controllers assuming the network is reliable. However, the network is subject to link failures. In this paper, we formulate a novel multi-objective SDN controller placement problem with the aim to minimize the switch-to-controller communication delay for both the cases without link failure and with single-link-failure. We propose an efficient metaheuristic-based Reliability-Aware and Latency-Oriented controller placement algorithm (RALO) for multi-objective multiple controller placements. The algorithm constructs an initial feasible solution by a greedy method with network partition, then repeatedly generates new solutions with variable neighborhood search. Once a new solution is generated, the algorithm decides whether to accept the new solution as a non-dominated solution to the problem and performs update operation on the Pareto optimal solution set. Meanwhile, to avoid falling into the local optimum, the algorithm also performs perturbation and destruction operations on the current solution. We finally conduct experiments through simulations on 8 real network topologies and two kinds of generated networks conforming to ER (Erdos–Renyi) random model and small-world model. Experimental results demonstrate that the proposed algorithm can achieve a competitive performance of switch-to-controller latencies in both the cases without link failure and with single-link failure, and the accumulated delay of primary and backup paths between the controllers and the switches. The Pareto optimal solution set provided by algorithm RALO allows network administrators with flexible choices to strike a trade-off between the switch-to-controller delay of primary and backup paths.