Leveraging load balance metrics to unravel the impact of multi-access edge computing locations on online dynamic network performance

Abstract

Telecommunication operators are increasingly relying on Network Function Virtualization (NFV) and Multi-access Edge Computing (MEC) to support emerging 5G/6G services, which demand ultralow latency and ultra-reliability. Employing NFV and MEC enable operators to deliver services through Service Function Chains (SFC) composed of Virtual Network Functions (VNFs) utilizing computing resources close to the end user. A critical challenge in this architecture is the efficient allocation of these resources and the strategic placement of MEC sites to host VNFs. This paper introduces, for the first time, a novel approach to efficiently determine where to locate MEC sites with the aim of optimizing dynamic performance. Instead of conducting time-consuming simulations to evaluate and compare each and every potential selection of MEC sites, we demonstrate that by quickly precomputing load balance metrics, such as the Jain fairness index (JFI), promising sets of sites can be identified. Our research shows that there is a statistically significant negative monotonic relationship between the precomputed JFI and the blocking probability when, during network operation, SFCs are dynamically established and released. Thus, by leveraging this fast identification method, network operators can focus their efforts, such as conducting detailed dynamic simulations (necessarily long and time-consuming since networks should operate with low or very low blocking ratios), solely on the most promising combinations. Therefore, this approach streamlines the process of determining the strategic location of MEC sites in a network, reducing the time required to plan and optimize the network configuration effectively.