On the optimal network for multicomputer: Torus or hypercube?
This paper examines the relative performance merits of the torus and hypercube when adaptive routing is used. The comparative analysis takes into account channel bandwidth constraints imposed by VLSI and multiple-chip technology. This study concludes that it is the hypercube which exhibits the superior performance, and therefore is a better candidate as a high-performance network for future multicomputers with adaptive routing.
Unable to display preview. Download preview PDF.
- 1.S. Abraham, Issues in the architecture of direct interconnection networks schemes for multiprocessors, Ph.D. thesis, Univ. of Illinois at Urbana-Champaign(1991).Google Scholar
- 3.R. Arlanskas, iPSC/2 system: A second generation hypercube, Proc. 3rd ACM Conf. Hypercube Concurrent Computers and Applications, Vol. 1 (1988).Google Scholar
- 4.Cray Research Inc., The Cray T3E scalable parallel processing system, on Cray’s web page at http://www.cray.com/PUBLIC/product-info/T3E/.Google Scholar
- 6.J. Duato, A New theory of deadlock-free adaptive routing in wormhole routing networks, IEEE Trans. Parallel & Distributed Systems, Vol. 4 (12) (1993) 320–331.Google Scholar
- 7.J. Duato, M.P. Malumbres, Optimal topology for distributed shared-memory multiprocessors: hypercubes again?, Proc. EuroPar’96, Lyon, France, (1996) 205–212.Google Scholar
- 8.R.E. Kessler, J.L. Schwarzmeier, CRAY T3D: A new dimension for Cray Research, in CompCon, Spring (1993) 176–182.Google Scholar
- 9.M. Ould-Khaoua, An Analytical Model of Duato’s Fully-Adaptive Routing Algorithm in k-Ary n-Cubes, to appear in the Proc. 27th Int. Conf. Parallel Processing, 1998.Google Scholar
- 10.C.L. Seitz, The Cosmic Cube, CACM, Vol. 28 (1985) 22–33.Google Scholar
- 11.C.L. Seitz, The hypercube communication chip, Dep. Comp. Sci., CalTech, Display File 5182:DF:85 (1985).Google Scholar