Dynamic Traffic Load Rebalancing for Hardware-accelerated 6G UPF Resilient Architecture

Recently, research on new 6G mobile network architectures has started to explore how to provide AI-enabling Infra using hardware accelerators in a cloud-native environment. User Plane Function (UPF) in mobile Core network, which handles multi-Tbps packet data from millions of mobile subscribers, could benefit from such offloading. However, UPF must be highly robust and resilient to provide the best service quality to end-users even under high traffic surge situations. We present a dynamic traffic load rebalancing mechanism in UPF using Integer Linear Programming (ILP) framework, to ensure even packet processing load allocation among Data Processing Units (DPUs). Our analysis and simulation results show the efficacy of the proposed scheme, demonstrating its potential for improved resiliency of hardware accelerated 6G UPF architecture.