Abstract
10G-passive optical network (XG-PON) provides an emerging potential technology to carry multiple services for the first - mile access network. However, it offers 0.3 ms propagation delay between the optical network Unit (ONU) and optical line terminal (OLT) resulting in inevitable increased queuing delay for the bandwidth grants to the ONUs. We present a reasonable and effective uplink dynamic bandwidth allocation method - demand forecasting DBA (DF-DBA) that predicts ONU's future demands by statistical modelling of the demand patterns and tends to fulfil the predicted demands just in time, which results in reduced delay. Results show that use of simple normal distribution in the DBA engine offers more appropriate bandwidth allocation with 14% decreased packet drop ratio (PDR) values than standard GigaPON access network (GIANT) DBA following request-grant cycle procedure. Proposed DBA when tested for all the T-CONTs, reduced 50% delay and offered 45% throughput improvement at higher traffic load. The unassigned bandwidth margin (UBM) is nearly ideal along with better QoS parameters results which validates that DF-DBA can practically improve the OLT performance. Further, this predictive DBA opens up a new horizon of research for researcher to use it for longer distances in PON (LR-PON) network with different statistical distributions for getting the possible best results in the future.