Forecast grant sizing for video delivery over passive optical networks

We investigate the use of video frame size forecasts to forecast video queue depths at the Optical Network Units (ONUs) on a Passive Optical Network (PON). These forecasted queue depths are used to allow the Optical Line Terminal (OLT) to provide exhaustive service to ONUs. Exhaustive service provides significantly lower queueing delay for video packets compared to Gated service. We propose a new video frame size forecasting method called Feed Forward Bandwidth Indication (FFBI) to be used in this application. We evaluate the use of FFBI and several other forecast methods and find that the use of FFBI in this application can provide statistically significant reductions in queueing delay and delay variation for video packets compared to other forecast methods.