QCN Based Dynamically Load Balancing: QCN Weighted Flow Queue Ranking

Ethernet has become the primary network protocol used due to its undeniable advantages such as low cost, high speeds or ease of management. Being a best effort protocol, the IEEE Data Center Bridging Task Group developed a series of Layer 2 enhancements including Quantized Congestion Notification (QCN) and enabled a loss less environment. In this paper we propose a method of dynamic network load balancing based on QCN, named QCN Weighted Flow Queue Ranking (QCN-WFQR). Several local and system wide congestion indicatives are computed related to a series of nodes within a network, based on which different decisions to balance the load in the congestion aware computer network can be taken. QCN-WFQR design is generic, so it can be used in traditional networks and in software defined networks as well, being the new approach in computer networking. The algorithm was implemented and tested in a discrete network simulator (NS3) where we have analyzed different profiles of congestion indicatives variations. Also, we propose one method of using QCN-WFQR in distributed and parallel file systems and one method of flows migrations in SDN achieving better load balanced networks.