Efficient collective data distribution in all-port wormhole-routed hypercubes

• Published in: Proceedings of the 1993 ACM/IEEE conference on Supercomputing pp. 792 - 801
• Main Authors: Robinson, D. F., Judd, D., McKinley, P. K., Cheng, B. H. C.
• Format: Conference Proceeding
• Language: English
• Published: New York, NY, USA ACM 01.12.1993; IEEE
• Series: ACM Conferences
• Subjects: Algorithm design and analysis; Computational modeling; Computer science; Computer systems organization > Architectures > Distributed architectures > Grid computing; Computer systems organization > Architectures > Parallel architectures; Computer systems organization > Architectures > Parallel architectures > Multicore architectures; Computer systems organization > Architectures > Serial architectures > Superscalar architectures; Computing methodologies > Modeling and simulation > Simulation types and techniques > Massively parallel and high-performance simulations; Concurrent computing; Hardware; Hypercubes; Multicast algorithms; Network topology; Networks > Network properties > Network structure; Routing; Software and its engineering > Software organization and properties > Software system structures > Distributed systems organizing principles > Grid computing; Theory of computation > Design and analysis of algorithms > Approximation algorithms analysis > Routing and network design problems; Unicast
• ISBN: 0818643404; 9780818643408
• ISSN: 1063-9535
• DOI: 10.1145/169627.169837

This paper addresses the problem of collective data distribution, specifically multicast, in wormhole-routed hypercubes. The system model allows a processor to send and receive data in all dimensions simultaneously. New theoretical results that characterize contention among messages in wormhole-routed hypercubes are developed and used to design new multicast routing algorithms. The algorithms are compared in terms of the number of steps required in each, their measured execution times when implemented on a relatively small-scale nCUBE-2, and their simulated execution times on larger hypercubes. The results indicate that significant performance improvement is possible when the multicast algorithm actively identifies and uses multiple ports in parallel.