Off-site protection against service function forwarder failures in NFV

Many internet applications in the emerging 5G/6G networks require ultra-reliable and low-latency communi-cations (URLLC) services. To deliver URLLC services flexibly and efficiently, network function virtualization (NFV) is employed to enable network slicing. NFV deploys service function chains (SFCs) consisting of service functions (SFs) and service function forwarders (SFFs) to deliver services. It is possible for SFFs in an SFC to fail when forwarding traffic to specified SF instances. Protecting SFF failures in NFV is challenging since multiple SF instances may fail simultaneously as a result of a single SFF failure. This work investigates how to protect the SF instances against the SFF failures while minimizing the cost of backup computing resources. We formulate a new problem called SFF-driven multi-instance failures and protection (SMFP), and prove its NP-hardness. We propose an efficient heuristic algorithm, namely, SFF-centralized resource optimization (FCRO), which is based on the proposed techniques of backup auxiliary transferring, backup cost-effectiveness selection, and adaptive fit backup. Our experimental results demonstrate that the proposed FCRO is effective and significantly outperforms the schemes directly extended from the existing work.