IQCN: Improved Quantized Congestion Notification for Ethernet Networks to provide stability and reduce data loss

Abstract

Congestion is a major problem for the data-processing Ethernet Networks, because it causes data loss and application failure in consequence. To this end, this kind of network needs to be suitably controlled in order to guarantee good performances. IEEE 802.1Qau standard is in progress to find an efficient congestion control mechanism, which may ensure a properly use of the available resources and avoid problems that cause this issue. In this context, several researches have contributed, however, none of them paid enough attention to resolve all issues. In this paper, we propose a new solution called IQCN as a contribution to IEEE 802.1Qau. This proposition is based on QCN Serial HAI and focuses on dealing with the congestion control encountered weaknesses. The main aim of the proposed mechanism is to provide stability and avoid/reduce data loss rate caused by congestion. To provide these purposes, IQCN used to reduce the signaling overhead cost and to predict the speed of incoming traffic to properly react face to the congestion issues. Our proposed mechanism overcomes and mitigates problems caused by congestion. Obtained simulation results using Omnet++ show that IQCN outperforms the QCN Serial HAI mechanism in terms of stability, fairness, overhead notification messages and data loss.