Two Algorithms for Fast Mitigating Multi-Congestion in Software Defined Networks

摘要

Although software defined networks (SDN) can achieve flexible network management, it lacks an efficient congestion control mechanism. Most existing researches focused on single link congestion scenario and has been inferior in the congestion mitigation time because of numerous flow operations between controllers and switches. Thus, in this paper, we address the problem of fast mitigating congestion (FMCP) in multiple link congestion (Multi-Congestion) scenario over SDN. We first formulate FMCP as a 0-1 nonlinear programming model, which aims to minimize the congestion mitigation cost and avoid congestion propagation. Then we propose two algorithms: one is an optimal algorithm based on Lagrange relaxation, called FMC-LR, which aims to transform the original optimization problem to several shortest path sub-problems corresponding to a new link weight; the other is a heuristic algorithm, called FMC-HE, which aims to reroute the congested flows traversing the maximum number of congested links to an alternative path with the minimum congestion mitigation cost until the congestion is mitigated. Extensive emulation results show that FMC-HE mitigates congestion faster than FMC-LR in small-scale networks, whereas FMC-LR outperforms FMC-HE in terms of success ratio and both algorithms can avoid congestion propagation