A Fault Diagnosis Method for Service Function Chaining with Adaptive Network Topology Changes

Abstract. To address the escalating operational costs and capital expenditures faced by operators due to the rigidity of dedicated network hardware equipment and difficulties in expanding new services in traditional network architectures, Network Function Virtualization (NFV) is introduced in 5G networks. With NFV, network functions can operate in the form of software on general-purpose servers, forming Virtual Network Functions (VNFs). Multiple VNFs are usually arranged in a specific order to form Service Function Chains (SFCs), providing users with flexible and efficient network services. To accommodate the changing user business requirements and utilize network resources more efficiently in NFV, SFCs need to be periodically migrated and reconfigured, leading to continuous dynamic changes in network topology. The mapping relationship between SFCs and the underlying physical network, along with the chained structure of SFCs, can result in both vertical and horizontal propagation of faults, thereby increasing the complexity of fault diagnosis. This paper proposes an adaptive fault diagnosis model generation algorithm for network topology changes to generate a corresponding fault diagnosis model based on the latest network topology. And the problem of vertical and horizontal fault propagation is then resolved based on the temporal nature of dynamic Bayesian networks. Experimental results show that the proposed fault diagnosis algorithm achieves an accuracy rate of 97.8%.