Quality of Experience from Cache Hierarchies: Caching for Adaptive Streaming over Information-Centric Networks
Video traffic is growing in dominance in today’s Internet, prompting new challenges in timely delivery of video content. This demand motivates the development of Dynamic Adaptive Streaming over HTTP (DASH), as a de facto paradigm for streaming service. DASH attempts to maximize the video consumers' Quality of Experience (QoE) under varying network conditions. However, as an application-level solution, DASH is struggling to cope with both low-latency delivery and massive service requests. The emerging Information-Centric Networking (ICN) architecture, with its ubiquitous cache hierarchies, promises a scalable adaptive streaming service. However, current ICN caching capability is aimed at generic traffic, and lacks the optimization needed for video-specific applications. As the performance of existing caching schemes is discovered to decline as consumers dynamically select content encoded at different bitrates, DASH evidently clashes with caching hierarchies, which reinforces the need to explore novel caching schemes that can incorporate bitrate adaptation, as a controlling mechanism in DASH, to serve this dominating traffic. In this dissertation, catering to adaptive video traffic is accomplished by caching mechanisms that emphasize various QoE measurements, such as video quality, bitrate oscillation, and rebuffering. Specifically, 1) the delivered video quality is enhanced by building a traffic model to estimate access delay under varying encoded bitrates. This delay is further used to optimize cache placement that causes the highest video throughput. 2) Bitrate oscillations and playback rebuffering are reduced due to our findings where bitrates should be prioritized in cache hierarchies. This is manifested in the design of a cache partitioning mechanism that safeguards cache capacity for specific bitrates. 3) Building on the premises of cache partitioning, the underlying dependency between cache hierarchies and bitrate adaptation is addressed for caching decisions that rely on instant video statistics. That is, instead of focusing on currently popular bitrates, bitrate adaptation predicts preferable rates in the future, and guides cache decisions that optimize the long-term performance under dynamic video requests. Extensive simulations demonstrate the effectiveness of proposed approaches in achieving significant improvements to QoE. These results suggest that cache hierarchies can better serve adaptive streaming when caching schemes capture, understand, and react to bitrate adaptation.