Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Interconnection networks: improved routing methods for Clos and Benes networks

Thesis/Dissertation ·
OSTI ID:6445478
Interconnection networks for use by multiprocessor systems are optimal in terms of area, but the routing algorithms used to set up these networks are suboptimal. This work improves routing on these networks using a four-fold approach. First, routing is examined in the matrix-decomposition domain; the new characteristic of partitioning is used to improve current algorithms in this class and to develop an algorithm that does not require backtracking. Second, edge coloring and matching on bipartite multigraphs are applied to routing on Clos networks. The regularity of the bipartite multigraph for routing applications is exploited to optimize a more-general edge-coloring algorithm. Third, double-coset decomposition is used to identify new properties of Benes networks and speed up routing for some permutations. Finally, the Tree-Connected Computer is used to parallelize the Looping Algorithm and reduce the setup time for Benes networks by a factor of O(lgN). It is also used to reduce the bandwidth requirements for the fastest current setup algorithm for the Complementary Benes network.
Research Organization:
Rensselaer Polytechnic Inst., Troy, NY (USA)
OSTI ID:
6445478
Country of Publication:
United States
Language:
English

Similar Records

Generalized schemes for access and alignment of data in parallel processors with self-routing interconnection networks
Journal Article · Thu Jan 31 23:00:00 EST 1991 · Journal of Parallel and Distributed Computing; (United States) · OSTI ID:5071023

Fault-tolerant multiprocessors with redundant-path interconnection networks
Journal Article · Mon Mar 31 23:00:00 EST 1986 · IEEE Trans. Comput.; (United States) · OSTI ID:5704786

Multicast routing in 3-stage Clos ATM switching networks
Journal Article · Mon Jan 31 23:00:00 EST 1994 · IEEE Transactions on Communications · OSTI ID:57335