Distrbuted Processing and Routing Strategy on the XMESH Topology for the Massively Parallel Computer Architecture 


Vol. 5,  No. 12, pp. 3109-3116, Dec.  1998
10.3745/KIPSTE.1998.5.12.3109


PDF
  Abstract

We propose the routing strategies on the XMESH topology for the massively parallel computer architecture with uniform message distribution and in the presence of link contention. And to confirm these strategies, the deflection routing algorithm and self-routing algorithm for the XMESH is used to simulate the maximum delay, the average delay, and the throughput in the presence of contention. For the deflection routing, we propose a new deflection criterion LD that the message with fewer optimal directional links has a higher priority, but for the messages with the same number of optimal directional links, the message with more deflections has a higher priority. And if the message that has the right to select link by the criterion has many optimal directional links, we adopt the strategy that the message has the preference for the diagonal link. The simulation results show that the maximum delay, the average delay, and the throughput using the proposed strategies are enhanced by about 58%, 70%, and 31%, respectively, compared with them using criterion A that the older message has a higher priority.

  Statistics


  Cite this article

[IEEE Style]

K. J. Jin and Y. S. Dae, "Distrbuted Processing and Routing Strategy on the XMESH Topology for the Massively Parallel Computer Architecture," The Transactions of the Korea Information Processing Society (1994 ~ 2000), vol. 5, no. 12, pp. 3109-3116, 1998. DOI: 10.3745/KIPSTE.1998.5.12.3109.

[ACM Style]

Kim Jong Jin and Youn Sung Dae. 1998. Distrbuted Processing and Routing Strategy on the XMESH Topology for the Massively Parallel Computer Architecture. The Transactions of the Korea Information Processing Society (1994 ~ 2000), 5, 12, (1998), 3109-3116. DOI: 10.3745/KIPSTE.1998.5.12.3109.