Ant Colony Optimization with Capacity-Aware Pheromone Models for Load-Balanced Service Placement in Fog Infrastructures

Abstract

Fog computing extends cloud capabilities toward the network edge to support latency-sensitive and bandwidth-intensive services arising from the proliferation of Internet of Things devices. In such infrastructures, services must be placed on heterogeneous fog nodes with limited capacity, while traffic demands vary over time and space. Unbalanced placement can lead to resource hot-spots, increased response times, and degraded quality of service. Traditional deterministic optimization methods often become intractable at the scale and dynamism of realistic fog deployments, motivating heuristic and metaheuristic approaches. This paper investigates ant colony optimization strategies for load-balanced service placement in fog infrastructures, focusing on capacity-aware pheromone models. The study considers a generic fog topology with constrained compute and bandwidth resources, a set of multi-tenant services with heterogeneous demands, and a traffic matrix describing the association between end-user regions and services. The placement problem is formulated as a linear mixed-integer model that jointly captures capacity constraints, routing decisions, and load-balance objectives. Upon this model, a family of ant colony algorithms is constructed in which pheromone values explicitly encode residual capacities, congestion indicators, and marginal load-balancing costs. Evaporation and reinforcement rules are designed to avoid convergence to placements that violate capacity limits or create persistent hot-spots. The resulting algorithms are discussed in terms of convergence behavior, structural properties of produced placements, and computational complexity. Numerical experiments on synthetic fog topologies illustrate how capacity-aware pheromone designs influence load distribution, path selection, and robustness under changing workload conditions.