A novel early DBA mechanism with prediction-based fair excessive bandwidth allocation scheme in EPON

In this paper, we propose a novel Early dynamic bandwidth allocation (E-DBA) mechanism incorporated with a prediction-based fair excessive bandwidth allocation (PFEBA) scheme in Ethernet passive optical networks (EPONs). The E-DBA mechanism can reduce the idle period in the traditional DBA mechanism. On the other hand, the PFEBA scheme can provide more accurate prediction to ensure the fairness of each ONU and improve the overall system performance. The proposed model makes prediction for different traffic classes according to the variation in traffic for each ONU in the EPON. The PFEBA scheme includes the unstable degree list, predictions made using linear estimation credit and the fair excessive bandwidth allocation scheme. The simulation results show that the proposed E-DBA mechanism with PFEBA scheme can improve the system performance of well-known DBA algorithms in terms of wasted bandwidth, wasted bandwidth improved percentage, downlink data available bandwidth, throughput, average end-to-end delay and average queue length, especially under high traffic load.