Congestion-Aware Algorithms for Service Function Chaining in Software-Defined Wide Area Networks.

In this paper, we address the service function chaining (SFC) problem for software-defined wide area networks (SD-WANs). Due to its NP-hardness, the service chaining problem is simplified by most existing works to apply to practical-size networks. However, in SD-WAN scenarios, these simplifications may lead to link congestion due to the heterogeneity of geographically distributed networks. Different from previous studies, we propose a congestion-aware algorithm, where more comprehensive and practical constraints are considered, and load balancing is introduced to reduce link congestion. First we formulate an integer linear programming model for exact solution, then the original problem is simplified and heuristic algorithms are designed to deploy the SFC requests in batches. Specifically, load balancing is implemented by updating the routing rules according to network congestion status, and data flows from different requests are dispersed in a probabilistic manner, which to our knowledge is first adopted for load balancing in SFC problems. Simulation results show that our algorithm can significantly reduce link congestion, thus improving the acceptance rate of SFC requests and obtaining higher benefits.