Multi-time-step Segment Routing based Traffic Engineering Leveraging Traffic Prediction

Based on the concept of source routing, Segment Routing (SR) allows the source or ingress node to inject a sequence of segment labels into the packet header and specify the routing path. Due to the routing flexibility, SR has been widely used to solve traffic engineering problems such as minimizing the maximum link utilization. However, most of the prior works only solve the problem in a single snapshot without considering network traffic dynamics, resulting in frequent traffic reroute. To cope with this problem, we focus on solving the segment routing based traffic engineering problem by taking into account the future traffic changes. We formulate the multi-time-step segment routing problem and leverage traffic prediction to extend the length of routing cycles. Due to the large search space of multi-time-step segment routing problem, we further propose a heuristic algorithm for incrementally recomputing the segment routing paths in sub-second. Through extensive experiments on real backbone network traffic datasets, we show that our proposal can achieve a near-optimal performance in term of maximum link utilization while significantly reducing the number of routing changes.