BGP-Based Clustering for Scalable and Reliable Gossip Broadcast

• Published in: Global Computing pp. 273 - 290
• Main Authors: Brahami, M., Eugster, P. Th, Guerraoui, R., Handurukande, S. B.
• Format: Book Chapter
• Language: English
• Published: Berlin, Heidelberg Springer Berlin Heidelberg
• Series: Lecture Notes in Computer Science
• Subjects: Broadcast streams; Content distribution; Fault-tolerance; Multicast; Peer-to-peer communications; Service overlay networks; System design
• ISBN: 9783540241010; 3540241019
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-540-31794-4_15

This paper presents a locality-based dissemination graph algorithm for scalable reliable broadcast. Our algorithm scales in terms of both network and memory usage. Processes only have “local knowledge” about each other. They organize themselves dynamically (right from the bootstrapping phase), according to join, leave or crash events, to form a locality-based dissemination graph. Broadcast messages can be disseminated using these graphs in large networks like the Internet, without relying on any special infrastructure or intermediate brokers. Roughly speaking, a dissemination graph consists of “non-crossing” (independent) trees that provide multiple paths between processes for improved broadcast efficiency and reliability. Each tree is constructed using BGP routing information about process “locality”. We convey the feasibility of the algorithm using both simulation and experimental results and describe an application of our algorithm for broadcasting information streams.