Affiliation:
1. Dublin City University, Ireland
Abstract
This paper proposes an adaptive unicast routing algorithm for large scale symmetric networks comprising 2 × 2 switch elements such as Beneš networks. This algorithm trades off the probability of blocking against algorithm execution time. Deterministic algorithms exploit the rearrangeability property of Beneš networks to ensure a zero blocking probability for unicast connections, at the expense of extensive computation. The authors’ algorithm makes its routing decisions depending on the status of each switching element at every stage of the network, hence the name adaptive routing. This method provides a low complexity solution, but with much better blocking performance than random routing algorithms. This paper presents simulation results for various input loads, demonstrating the tradeoffs involved.
Reference30 articles.
1. The Looping Algorithm Extended to Base 2<sup>t</sup>Rearrangeable Switching Networks
2. Batcher, K. E. (1976). The flip network in STARAN. In Proceedings of the International Conference on Parallel Processing (pp. 65-71)
3. Beetem, J., Denneau, M., & Weingarten, D. (1985). The GF11 supercomputer. In Proceedings of the 12th annual international symposium on Computer architecture (ISCA ’85).
4. Beizer, B. (1968). The Analysis and Synthesis of Signal Switching Network. In Proceedings of the Symposium on Mathematical Theory of Automata, NY (pp. 563-576)