Abstract
HTTP Adaptive Streaming (HAS) is an adaptive bitrate streaming technique which is able to adapt to the network conditions using conventional HTTP web servers. An HAS player periodically requests pre-encoded video chunks by sending an HTTP GET message. When the downloading a video chunk is finished, the player estimates the network bandwidth by calculating the goodput and adjusts the video quality based on its estimates. However, the bandwidth estimation in application layer is pretty inaccurate due to its architectural limitation. We show that inaccurate bandwidth estimation in rate adaptation may incur serious rate oscillations which is poor quality-of-experience for users. In this paper, we propose a buffer-based rate adaptation scheme which eliminates the bandwidth estimation step in rate adaptation to provide a smooth playback of HTTP-based streaming. We evaluate the performance of the HAS player implemented in the ns-3 network simulator. Our simulation results show that the proposed scheme significantly improves the stability by replacing bandwidth estimation with buffer occupancy estimation.