Ant algorithm based load balancing for network sessions

This paper presents a multiple-ant-colony load balancing algorithm (antBalance) for the resources allocation within network sessions. Combined ant colony algorithm with dynamic load balancing in the network traffic engineering, the algorithm further extends the basic ant colony to multiple ones with dynamic pheromone release design. By influencing the ant colony pheromone with each other, each ant in one ant colony not only strengthens the pheromone of the same colony, but also weakens that of the other colonies in the paths. antBalance tries to allocate ants to different paths fairly to avoid traffic congestion in some certain paths. Besides analyzing the effective control of the pheromone, compared with basic ant algorithm, simulation experiments illustrate that antBalance achieves better resources allocation performance in terms of path bandwidth utilization, session delay and session packet loss.