A framework for multicast video streaming over IP networks

• Published in: Journal of network and computer applications Vol. 26; no. 3; pp. 273 - 289
• Main Authors: Yeo, C.K., Lee, B.S., Er, M.H.
• Format: Journal Article
• Language: English
• Published: London Elsevier Ltd 2003; Elsevier
• Subjects: Applied sciences; Exact sciences and technology; Multicast; Overlay; Streaming network; Systems, networks and services of telecommunications; Telecommunications; Telecommunications and information theory; Transmission and modulation (techniques and equipments); Multicast; Overlay; Streaming network; Performance evaluation; Tree structured method; Internet protocol; Telecommunication network; Video signal; Service quality; Information transmission
• ISSN: 1084-8045; 1095-8592
• DOI: 10.1016/S1084-8045(03)00004-3

A self-configurable and self-improving application-level framework is proposed for the streaming of video data across heterogeneous networks. The framework comprises a dynamic tree of clients which also act as relays to distribute video downstream and monitor network conditions, a lightweight server to manage the tree and a gossip-spiral mechanism to improve the tree robustness. The framework is loosely coupled such that failure of the lightweight server will not impact the data distribution functionality of the tree. The motivation of such an application-level approach which does not require router intervention, stems from the fact that IP-multicast is not enabled prevalently in most parts of the Internet. Streaming applications thus have to depend on a network of servers, relays and static tunnels instead of the more efficient multicast for data distribution across the Internet. Such conventional static approach is less efficient than the proposed dynamic framework as the tree is configured on the fly taking prevailing network conditions into account. Moreover, it is able to self-improve and self-recover through the process of a client switching to another parent for a better QoS. The performance of the framework has been evaluated against unicast and multicast and both experimental and simulation results show that the framework is superior to unicast in performance. In fact, simulation results show that the framework is able to track ideal multicast (minimum spanning tree) in performance under non-congested network conditions.