Towards content delivery optimization in future wireless networks

The last few years have seen an enormous growth in the usage of smart phones and this trend is expected to continue in the near future with more processing power and larger form factors. This explosive growth has contributed to an unprecedented adoption of mobile data services and increase in data usage. The demand is also driving the development of next generation networks to support novel data hungry and user-experience focussed applications such as virtual reality video. Global mobile data traffic is increasing at an unprecedented rate and video traffic alone currently constitutes about 50% of the total traffic and is predicted to grow up to 70% of the total traffic by 2021 as user preferences are shifting towards more video based applications relative to browsing. The explosion of traffic associated with video content poses significant challenges for mobile content provision. While, on the one hand, mobile video traffic surge is forecasted to require significant investments in bandwidth acquisition and infrastructure deployment and roll-out, on the other hand, users are not likely to be willing to pay significantly more than today. Operators would like to charge the content providers for revenue. Furthermore, user expectations for high quality video is constantly increasing.;Our goal in this thesis is to find out how to best deliver video content without additional investment from providers while still maintaining similar user Quality of Experience (QoE) as of today. We define QoE in terms of overall perception of service which include direct factors such as perceived quality of video content and indirect factors such as battery lifetime. We address the problem using three methods : (1) context-aware opportunistic pre-fetching of content, (2) video content optimization based on user perception and (3) video content delivery with personalized quality. Our results show that opportunistic prepositioning of data objects which are likely to be consumed by users (pre-fetching) utilizing contextual parameters such as location, high data rates in an over-the-top fashion can reduce the energy consumption of transmission for video streaming by up to 40%. Next, by adapting the video based on user perception, where a video is broken into shorter segments which could be delivered in a satisfying resolution, we were able to show that file sizes of video can be reduced by up to 60% while achieving similar user perceived quality as the original video in fixed resolution. This technique allows us to create video content in varying perceptual qualities. Finally, we show that by grouping users based on similarities in viewing history for predicting and delivering a specific video quality (personalized) for individual users, we can reduce the overall video traffic in the network by up to 50%.