Balls and bins distributed load balancing algorithm for channel allocation

In this paper we validate through simulations the convergence properties of our earlier proposed balls and bins channel allocation algorithm. The algorithm supports dynamic channel allocation based on local link quality information and does not require exchange of network information. Each radio evaluates periodically a cost function related to the desired QoS. Based on the calculated cost the radio may change the channel with probability proportional to the excess of the current cost over a cost threshold set by the user application. Earlier theoretical work showed the algorithm has low complexity and excellent convergence properties. In this paper we show that radio network simulations confirm the results. In terms of aggregate network throughput the algorithm offers on average modest gains. The most remarkable improvement that the proposed algorithm offers is in terms of fairness. The results show that the worst link throughput in the network can be improved up to 70%. The algorithm guarantees a certain minimum performance to all the links in the network that is determined by the maximum cost set. If all the transmitters have the same cost threshold and there exists a suitable link allocation for the cost threshold set then fairness in the convergence state is guaranteed.