Joint Resource Provisioning and Allocation for Service Function Chaining in Software-Defined Wide Area Networks

The software-defined wide area network (SD-WAN) has emerged as a promising architecture for the next-generation WANs, where the Internet service provider (ISP) leases network resources from the infrastructure provider (InP) and offers various applications through service function chaining. Taking the perspective of an ISP, this study focuses on two closely coupled issues. 1) Resource Allocation (RA): In bandwidth-constrained SD-WANs, network resources need to be flexibly allocated according to link congestion status. 2) Resource Provisioning (RP): Due to service dynamics, ISPs need to adjust their resource leasing schemes in response to request variations. The joint consideration of RA and RP enables bidirectional adaptation between the network and services. In this work, an ILP problem is formulated to unify RA and RP within one framework. For RA, a congestion-aware heuristic algorithm is proposed to achieve load balancing through probabilistic dispersion of data flows across different requests. For RP, a Bayesian Optimization-based algorithm is designed to facilitate resource capacity planning for ISPs. Simulations demonstrate that the proposed methods can reduce request rejection by 50% in bandwidth-constrained networks, improve resource utilization efficiency, enhance the ISP’s benefit, and mitigate performance fluctuations caused by service dynamics.