Cooperative Ant Colony Optimization for Multi-Commodity Flow and Congestion Management in Software-Defined Networks

Abstract

Software-defined networking decouples the control plane from the data plane and exposes a global view of the network to logically centralized controllers. This programmability enables fine-grained traffic engineering and flow-level optimization that are difficult to realize in traditional distributed routing architectures. At the same time, the growth of heterogeneous traffic and multi-tenant services in data center and wide-area deployments introduces multi-commodity flow patterns with stringent performance and isolation requirements. Congestion management in such environments requires models and algorithms that can exploit centralized visibility while remaining scalable with respect to the number of flows, links, and control epochs. Exact mathematical programming approaches can express these requirements but often become computationally expensive for large networks and short reconfiguration intervals. Metaheuristic approaches can provide approximate solutions within practical time budgets, yet they must be carefully adapted to the structural properties of multi-commodity flow constraints and controller architectures. This paper investigates a cooperative ant colony optimization framework for multi-commodity flow routing and congestion management in software-defined networks. The study combines a linear programming formulation of the traffic engineering objective with a multi-colony ant-based search process that is guided by link-level and path-level pheromone information. Emphasis is placed on the interaction between the linear model and the heuristic components, on strategies for cooperative information sharing among colonies, and on the analysis of congestion-aware pheromone updates under controller resource constraints.