Design and analysis of the Dual-Torus Network
- 31 Downloads
In this paper, we propose a new topology called theDual Torus Network (DTN) which is constructed by adding interleaved edges to a torus. The DTN has many advantages over meshes and tori such as better extendibility, smaller diameter, higher bisection width, and robust link connectivity. The most important property of the DTN is that it can be partitioned into sub-tori of different sizes. This is not possible for mesh and torus-based systems.
The DTN is investigated with respect to allocation, embedding, and fault-tolerant embedding. It is shown that the sub-torus allocation problem in the DTN reduces to the sub-mesh allocation problem in the torus. With respect to embedding, it is shown that a topology that can be embedded into a mesh with dilation δ can also be embedded into the DTN with less dilation. In fault-tolerant embedding, a fault-tolerant embedding method based on rotation, column insertion, and column skip is proposed. This method can embed any rectangular grid into its optimal square DTN when the number of faulty nodes is fewer than the number of unused nodes. In conclusion, the DTN is a scalable topology well-suited for massively parallel computation.
KeywordsDual Torus Network (DTN) l-Complete Torus Mesh Embedding Partitionable Torus System Sub-torus Allocation Faulttolerance
Unable to display preview. Download preview PDF.
- 3).Chuang, P.-J. and Tzeng, N.-F., “An efficiet submesh allocation strategy for mesh computer systems,” inProc. 11th International Conf. on Distributed Computing Systems, pp. 256–262, 1991.Google Scholar
- 5).Ding, J. and Bhuyan, L. N., “An adaptive submesh allocation strategy for two-dimensional mesh connectd systems,” inProc. 1993 International Conf. on Parallel Processing, vol. II, pp. 22–28, 1993.Google Scholar
- 7).Huang, S.-H.S., Liu, H. and Verma, R. M., “On embeddings of rectangles into optimal squares,” inInternational Conf. on Parallel Processing, vol. III, pp. 73–76, 1993.Google Scholar
- 8).Huang, S.-H. S., Liu, H. and Verma, R. M., “A new combinatorial approach to optimal embedding of rectangles,” in8th International Parallel Processing Symp., pp. 715–722, 1994.Google Scholar
- 9).Johnson, B. W.,Design and Analysis of Fault Tolerant Digital Systems, Addison-Wesley, 1989.Google Scholar
- 10).Kumar, V., et al.,Introduction to Parallel Computing: Design and Analysis of Algorithms, The Bejamin/Cumming Publishing Company, Inc, 1994.Google Scholar
- 11).Leiserson, C. E., “Area-efficient graph layouts(for VLSI),” inProc. 21st IEEE Symp. on Foundations of Computer Science, pp. 270–281, 1980.Google Scholar
- 12).Li, K. and Cheng, K. H., “A two dimensional buddy system for dynamic resource allocation in a partitionable mesh connected system,” inProc. of ACM Computer Science Conf., pp. 22–28, 1990.Google Scholar
- 13).Li, K. and Cheng, K. H., “Complexity of resource allocation and job scheduling problem in partitionable mesh connected systems,” inFirst Ann. IEEE Symp. Parallel and Distributed Processing, pp. 358–365, 1989.Google Scholar
- 14).Liu, T., Huang, W.-K., Lombardi, F. and Bhuyan, L. N., “A submesh allocation scheme for mesh-connected multiprocessor systems,” inProc. 1995 International Conf. on Parallel Processing, vol. II, pp. 159–163, 1995.Google Scholar
- 15).Methern, R. and Hwang, G., “Embedding rectangular grids into square grids with dilation two,”IEEE Trans. on Computers, no. 12, pp. 1446–1445, 1990.Google Scholar
- 16).Sharma, D. D. and Pradhan, D. K., “A fast and efficient strategy for submesh allocation in mesh connected parallel computers,” inProc. of 1993 The 5th IEEE Symp. on Parallel and Distributed Processing, pp. 682–689, 1993.Google Scholar
- 17).Shen, X., Liang, W. and Hu, Q., “Embedding between 2-d meshes of the same size,” inProc. of the Fifth IEEE Symp. on Parallel and Distributed Processing, pp. 712–719, 1993.Google Scholar
- 18).Thompson, C. D., “A Complexity Theory for VLSI,”Ph.D. thesis, Carnegie-Mellon University, Pittsburg, PA, Jun. 1980.Google Scholar
- 19).Yoo, B. S., Das, C. R. and Yu, C., “Processor management techniques for mesh connected multiprocessors,” inProc. 1995 International Conf. on Parallel Processing, vol. II, pp. 105–112, 1995.Google Scholar
- 20).Yu, C., Mohapatra, P., Das, C. R. and Kim, J., “A lazy scheduling scheme for improving hypercube performance,” inProc. 1993 International Conf. on Parallel Processing, vol. 1, pp. 110–117, 1993.Google Scholar
- 21).Zhu, Y., “Efficient processor allocation strategies for mesh-connected parallel computers,”Journal of Parallel and Distributed Computing, pp. 328–337, 1992.Google Scholar