Multi-Criteria Optimization for Service Placement and Migration in Collaborative Edge Computing

The rapid growth of latency-sensitive applications in edge computing requires efficient strategies for service placement and migration. However, existing approaches often prioritize a single objective (e.g., latency) while neglecting the associated trade-offs in terms of energy, cost, and resource utilization. In this study, we examine the combined issues of service placement and migration within a Collaborative Edge Computing (CEC) environment. In this environment, conflicting objectives such as end-to-end latency, energy consumption, and migration cost must be balanced under capacity (Central Processing Unit (CPU), memory, and bandwidth) and Quality of Service (QoS) constraints such as maximum delay. We propose a multi-objective formulation that addresses these criteria simultaneously, and we introduce a distributed evolutionary optimization method based on Non-dominated Sorting Genetic Algorithm-II (NSGA-II) in an island model. We evaluated our method using real-life scenarios involving cloud-region-micro-edge topologies, heterogeneous loads, and differentiated deadlines, and compared it with a centralized NSGA-II approach and a weighted sum approach. Our results reveal significantly broader Pareto coverage and greater solution diversity. This results in larger hypervolumes and more favorable trade-offs at the extremes of the front (minimum latency, minimum energy, or contained migration). Two-dimensional projections confirm that the proposed method outperforms the baselines on several objective planes. Spatial visualization of the mapping reveals local service anchoring that improves latency without increasing migration costs.