Control-Plane Load Balancing and Auto-Scaling in 5G and Beyond Networks

In this paper, we study an auto-scaling and load balancing approach, called ScaleLB, in a 5G core control plane with an aim to minimize control plane latency while maximizing the resource utilization of active network functions. Specifically, we focus on the latency involved in gNB association and end-user authentication through the access and mobility management function (AMF) in the core network. ScaleLB considers two approaches for auto-scaling and load balancing – proactive and reactive – compared to the static approach. To minimize control plane latency, the proactive approach scales the number of AMFs based on a predefined threshold irrespective of future service requests. In contrast, on receiving new requests, the reactive approach scales the number of AMFs when the existing AMFs are fully utilized. We develop a prototype of the system using open-source software tools. Experiment results present the trade-off between latency and the number of active AMFs with proactive and reactive schemes while comparing them with the static approach. Furthermore, the proactive approach yields competitive performance with the static approach when the threshold value for scaling up is carefully considered.