Thanks to their increasing sophistication and popularity, mobile devices, in the form of smartphones and tablets, have become the fastest growing contributors to Internet traffic. Indeed, smartphones are projected to account for 50% of global Internet traffic by 2017, with the share of mobile video increasing to about 40% of total Internet traffic. As users embrace Internet streaming of video, several studies have found that a small decrease in video quality leads to a substantial increase in viewer abandonment and disengagement rates. To handle the explosive growth in video traffic, Adaptive HTTP streaming, which exploits the prevalence of commodity web servers and content distribution networks, has emerged as the key technology for delivering video to end users. Although a number of systems have been proposed for HTTP video streaming in traditional environments and for fixed clients, existing platforms for video streaming on mobile devices are still in their infancy and do not address the additional challenges often experienced by mobile clients: high fluctuations in network conditions, heterogeneous networking interfaces, multiple form-factors, and limited battery life.
In this dissertation, we propose a number of solutions for improving the Quality of Experience of HTTP video streaming on mobile devices. We begin by evaluating the performance of several existing video quality adaptation schemes when deployed on mobile platforms. Through experiments with smartphones in wide-area environments, we assemble several key findings. First, we show that the high fluctuations in network throughput on cellular and Wi-Fi networks impose significant challenges for efficiently architecting the video adaptation scheme. Second, we find significant differences between the performance of the current state-of-the-art schemes in controlled experimental settings and their performance in mobile settings on key quality metrics such as inefficiency, instability, rebuffering ratio, and startup latency. We also find noticeable differences in the behavior of the schemes under Wi-Fi and cellular network access, with most of the schemes performing worse when the network access is cellular. Given these observations, we hypothesize on the possible causes of these inefficiencies. We also identify the best practices of existing schemes and key insights from experimental results that can serve as foundations for addressing many of the limitations.
Armed with these measurement-driven insights, we propose a novel video quality adaptation scheme, called MASS, which is more robust to the vagaries of the wireless networking conditions. We implement and evaluate our solution on commodity Android smartphones, and demonstrate significant performance gains over existing schemes. To further improve the streaming experience, we introduce an extension to HTTP video streaming that leverages the synergy between social network participation and video streaming to optimize end-user Quality of Experience. Our system, called SDASH, integrates and applies well-known concepts such as cooperative caching, prefetching, and P2P streaming for reducing bitrate fluctuations and optimizing the viewing experience. Finally, we develop a general infrastructure for constructing temporally and spatially localized P2P communities of mobile devices sharing similar interests. The platform enables on-demand cooperation among mobile clients based on device context and client preferences. We use a concrete implementation of the mobile P2P infrastructure for evaluating the performance of SDASH.
This dissertation addresses the challenges facing Adaptive HTTP Streaming under mobile networking conditions. Through experimentation with commodity mobile devices, we show that the proposed techniques for bitrate adaptation and cooperative streaming can significantly improve the video viewing experience.
| Identifer | oai:union.ndltd.org:GATECH/oai:smartech.gatech.edu:1853/53423 |
| Date | 08 June 2015 |
| Creators | Yusuf, Lateef |
| Contributors | Ramachandran, Umakishore |
| Publisher | Georgia Institute of Technology |
| Source Sets | Georgia Tech Electronic Thesis and Dissertation Archive |
| Language | en_US |
| Detected Language | English |
| Type | Dissertation |
| Format | application/pdf |
Page generated in 0.0854 seconds