Although, wireless solutions continue to be a dominant enabling technology in the future backhaul segment, they are susceptible to weather disturbances that may substantially degrade network throughput, or delay, compromising the 5G requirements. These effects can be alleviated by centralized rerouting realized by software defined networking (SDN) architecture. However, careless frequent reconfigurations may lead to inconsistencies in network states due to asynchrony between different switches, which may create congestion and limit the gain of frequent rerouting. In this paper, we focus on the rerouting process during rain disturbance considering the minimum total congestion imposed during the update of routing tables as a switching cost. At each time sample, the central controller has the possibility to adopt the optimal routes at a switching cost or to keep using previous routes at the expense of a throughput loss due to route sub- optimality. To find optimal solutions with minimal data loss in a static scenario, we formulate a dynamic programming problem that utilizes perfect knowledge of the rain attenuation for the whole rain period (off-line policy with full knowledge). For dynamic scenarios where the future rain attenuation data cannot be known, we propose an online consistency-aware rerouting algorithm, called optimal control action with prediction (OCAP), which uses the temporal correlation of rain fading to estimate the future rain attenuation. Simulation results on synthetic and real networks validate the efficiency of our OCAP algorithm, substantially reducing congestion and increasing network throughput with a fewer number of rerouting actions compared to benchmarks approaches.
IEEE Communications Society, 2017.