Abstract
Clients of streaming services have very diverse service expectations, demanding for differentiated services provisioning from the streaming servers. In this paper, we present a differentiated bandwidth allocation strategy enabled by real-time video adaptation technologies. It aims to provide differentiated streaming bit rates from the server to different client classes in proportion to their pre-specified differentiation weights, independent of the class loads. The allocation strategy is based on the predicted arrival rate of each class and the measured bandwidth release rate of the server. We propose a feedback queue technique, which utilizes the information of the backlogged requests in listen queues to accurately estimate the arrival rates, to improve the differentiation robustness. Simulation results show that the allocation strategy with a FCFS/FF request scheduler can meet the objective of proportional streaming bit rate differentiation in both short and long timescales and greatly enhance the service availability and maintain low queueing-delay when the streaming server load is high.