Data plane assisted state replication with Network Function Virtualization

Modern 5G networks are capable of providing ultra-low latency and highly scalable network services by employing modern networking paradigms such as Software Defined Networking (SDN) and Network Function Virtualization (NFV). The latter enables performance-critical network applications to be run in a distributed fashion directly inside the infrastructure. Being distributed, those applications rely on sophisticated state replication algorithms to synchronize states among each other. Nevertheless, current implementations of such algorithms do not fully exploit the potential of the modern infrastructures, thus leading to sub-optimal performance. In this paper, we propose STARE, a novel state replication system tailored for 5G networks. At its core, STARE exploits stateful SDN to offload replication-related processes to the data plane, ultimately leading to reduced communication delays and processing overhead for VNFs. We provide a detailed description of the STARE architecture alongside a publicly-available P4- based implementation. Furthermore, our evaluation shows that STARE is capable of scaling to big networks while introducing low overhead in the network.