Using Exponential and Logarithmic Thresholds to Optimize Scalable Video Traffic over Passive Optical Networks

Demand for internet services is rising quickly with the recent growing influx of internet-connected devices. Networks are continually being optimized or upgraded to accommodate this traffic with Passive Optical Networks (PONs) being a popular choice for replacing dated infrastructure due to their high performance to cost ratio. However, video traffic makes up an increasingly large portion of global IP traffic causing optimization to be difficult. This difficulty stems from the high coefficient of variability of compressed video traffic and is then compounded by the growing use of PONs which commonly employ a star topology thereby mandating the use of a centralized MAC protocol. Video traffic over PONs can be optimized by changing a video stream's bit-rate during transmission thus adapting to the bandwidth available to a PON at any point in time. This is accomplished by dropping a variable number of Medium Grain Scalable (MGS) H.264 AVC video enhancement layers. In this work two novel packet dropping schemes are proposed, one employing a logarithmic dropping threshold and the other employing an exponential dropping threshold. It is shown that though the performance improvements gained from dropping all of the scalable video layers is small, the most efficient way to do so is through employing the logarithmic threshold dropping scheme which achieved a 146.9% improvement over the base layer for video delay jitter easily outperforming the 89.8% and 44.5% improvement for linear and exponential dropping respectively. The benefits of logarithmic dropping over linear and exponential dropping are reinforced in the data queueing delay and video queueing delay performance metrics.