Edge server placement and allocation optimization: a tradeoff for enhanced performance

Considering the expansion of the Internet of Things (IoT) and the volume of data and user requests, Mobile Edge Computing (MEC) is considered a novel and efficient solution that puts decentralized servers at the network’s edge. This has the effect of lowering bandwidth demand and transmission latency. Optimal edge server placement and allocation, as the first stage of MEC, can improve end-user service quality, edge computing system utility, and cost and energy consumption. The majority of previous edge server placement studies have employed only one objective or developed a fitness function by the weighted sum method for optimization. Usually, using a single optimization objective without considering other objectives cannot yield the desired results for a problem with a multi-objective design. On the other hand, assigning weights to objectives can lead to losing optimal points in non-convex problems and selecting improper weights. Therefore, in this paper, we propose a multi-objective solution for the positioning and allocation of edge servers for MEC services based on the NSGA-II algorithm. In this regard, we identify two workload variance and latency reduction objectives with extensive evaluations. The experimental evaluation of the results using real-world data reveals that solutions based on the NSGA-II yield superior convergence and diversity of Pareto front points compared to Multi-Objective Particle Swarm Optimization (MOPSO), Multi-Objective Biogeography Based Optimization (MOBBO), and Adaptive Weighted Sum Method (AWSM). Additionally, it effectively mitigates workload variance on servers and exhibits an average latency reduction of 8.79% in comparison to the adaptive weighted-sum approach, 9.19% in comparison to MOPSO, and 0.28% in comparison to MOBBO.