Cache-friendly Rate Adaptation for Dynamic Adaptive Streaming over HTTP (DASH)

Abstract

The Internet in the recent years has seen a rapidly growing demand for multimedia content streaming. In order to deliver the streaming services to every corner of the Internet, HTTP streaming technologies have been widely adopted to replace the traditional RTSP/RTP streaming, due to the fact that HTTP streaming can avoid the issues arising from firewalls and NATs. Among the popular HTTP streaming technologies, Dynamic Adaptive Streaming over HTTP (DASH) has drawn the spotlights very recently.

In this thesis, we make comprehensive studies on the HTTP streaming technologies and specifically on DASH. By investigating various aspects of the DASH technology together with its underlying protocol and CDN infrastructures, we are able to identify a major problem posed by CDN caches, which still limits the performance of DASH. After understanding the advantages and drawbacks of the solutions proposed by other researchers, we have devised a unique client side rate adaptation algorithm, hoping to improve the performance of DASH in CDN networks, with a simple solution. Multiple experiments are designed and conducted to test our proposed algorithm. By studying the experiment results, we reveal how DASH performs under various network conditions, and at the same time make some conclusion on the design principles of a DASH client rate adaptation algorithm. Apart from the related studies and the algorithm proposal, some criticism is also made at the end of this thesis, as part of our DASH research conclusion.