ISBN 978-3-8439-2836-6

978-3-8439-2836-6, Reihe Informatik

Julius Rückert
Large-scale Live Video Streaming over the Internet - Efficient and Flexible Content Delivery Using Network and Application-Layer Mechanisms

276 Seiten, Dissertation Technische Universität Darmstadt (2016), Hardcover, B5

Zusammenfassung / Abstract

Video delivery over the Internet continues to dominate network traffic all over the world. After the success of video-on-demand services over the last decade, recently a notable shift of live video to the Internet can be observed. To enable future live streaming scenarios with hundreds of millions of concurrent users, new approaches are required to drastically increase the efficiency and flexibility of existing content delivery mechanisms.

To this end, this doctoral thesis identifies two essential areas of research that have the potential to jointly address these future needs. General and research-area-specific requirements for delivery mechanisms are derived, a conceptual architecture is defined for the integration of mechanisms from the two areas, and two novel delivery mechanisms are proposed, each focusing on one of the identified research areas.

The first contribution is the design of Software-Defined Multicast (SDM), a cross-layer mechanism based on software-defined networking. SDM introduces network-layer multicast support for the delivery of over-the-top live video streams inside Internet service provider networks. The mechanism is shown to be efficient as well as flexible and illustrates how Internet service providers could become active players in future content delivery scenarios.

The second contribution focuses on overlay-based concepts that enable scaling live content delivery in parts of the Internet where network-layer multicast support is unavailable. A novel peer-to-peer-based hybrid streaming mechanism called Topology-Optimized Transit (TopT) is proposed. TopT effectively combines existing and new streaming mechanisms into a common architectural framework to enable transitions between mechanisms and adapt to highly dynamic streaming workloads.