Towards Optimal Disaster Recovery in Backbone Networks

Fast recovering backbone networks from failures caused by unexpected disasters is critical to enhance the Internet resilience. The major challenge is fast switchover traffic from a faulty path to an alternative one. In this paper, we present an approach toward achieving a zero switchover time (i.e., optimal) by using two emerging technologies: Multipath TCP (MPTCP) and Software-Defined Networking (SDN). The conceptual idea is keeping multiple concurrent paths in an end-to-end communication, which is continuously alive unless all the paths get failure. Besides that, when a disaster occurs, traffic engineering is adopted to quickly balance the traffic on available paths. In order to establishes the concurrent paths, MPTCP divides an application´s byte stream into multiple TCP subflows, each of which is forwarded in an end-to-end route. The routes are normally determined by IP routing protocols, such as OSPF or BGP, which have two disadvantages. The first one is the protocols always take long time to reestablish new routes in disasters (i.e., long convergence time). The second one is they are originally designed without traffic engineering. On the other hand, SDN enables the use of central controllers with global view of dynamic network state. In addition, the controllers can remotely controls both direction and capacity of traffic flows in many flexible ways (i.e., using OpenFlow). Therefore, SDN potentially solves the disadvantages of IP routing protocols.