The explosion of video traffic in the Internet is mainly fueled by two types of applications: video broadcasting (live streaming) and video-on-demand (VOD). Furthermore, high-resolution, free-viewpoint and multiview videos have brought new market opportunities. Distributing these videos to large number of distributed users presents new challenges and calls for novel design in both network and streaming algorithms. In this thesis, we study the optimization of overlay network so as to cost-effectively support such next-generation video services. This includes overlay algorithm design and cost optimization for live broadcasting, and optimal replication for VOD.

First, we study delay optimization for high bitrate overlay video broadcasting. We propose an efficient push-based multi-t...[ Read more ]

The explosion of video traffic in the Internet is mainly fueled by two types of applications: video broadcasting (live streaming) and video-on-demand (VOD). Furthermore, high-resolution, free-viewpoint and multiview videos have brought new market opportunities. Distributing these videos to large number of distributed users presents new challenges and calls for novel design in both network and streaming algorithms. In this thesis, we study the optimization of overlay network so as to cost-effectively support such next-generation video services. This includes overlay algorithm design and cost optimization for live broadcasting, and optimal replication for VOD.

First, we study delay optimization for high bitrate overlay video broadcasting. We propose an efficient push-based multi-tree algorithm to achieve a low-delay high-bandwidth overlay backbone. The algorithm, termed FastMesh, constructs the overlay accommodating asymmetric and diverse uplink bandwidth. FastMesh continuously improves delay based on existing nodes. We further examine the feasibility of multi-Mbps streaming through a global measurement study using global experiments.

Second, we study large-scale multiview and free-viewpoint video broadcasting, where videos from different viewpoints of the same 3D scene are captured by multiple cameras. Users may select at will different camera angles, which are synthesized using texture and depth videos of the sandwiched camera views or the so-called anchor views. We design a collaborative broadcasting network of free viewpoint videos, where users may interactively pull and co-operatively share streams of different anchor views. We optimize the anchor views allocated to users so as to minimize the overall streaming cost.

Third, we study optimal replication for interactive multiview video on demand (VOD). We propose a redundant coding structure that facilitates interactive view switching, trading off storage with transmission rate. Using the coding structure, we next propose a content replication strategy that takes advantage of indirect hit to lower view-switching cost: in the event that the exact requested view is not available locally, the local server can fetch a different but correlated view from the other servers, so that the remote repository only needs to supply the pre-encoded view differential.