Broadcast communication delay metric for the iPSC/2 and iPSC/860 hypercubes

ABSTRACT

When determining the granularity of a program to be executed in parallel, it is important to have valid information on the cost of communication for the system on which the program is to be run. Previous studies have measured the node-to-node communication performance on the Intel Hypercube systems. This work measures the performance of multicast communication on the Intel iPSC/2 and iPSC/860 hypercubes and derives equations modelling the cost of broadcast communication.

REFERENCES

• 1: J. Baxter and J. Patel, "The LAST Algorithm: a Heuristic- Based Static Task Allocation Algorithm", .Proc. 1989 ICPP, pp. 217-222.
• 2: A. Beguelin and D.J. Vasicek, "Communication Between Nodes in a Hypercube", Hypercube Multiprocessors 1987, SIAM, Philadelphia, Pa, 1987.
• 3: D. Bergmark, et al, "On the Performance of the FPS T- Series Hypercube", Hypercube Multiprocessors 1987, SIAM, Philadelphia, Pa, 1987.
• 4: Y. Birk, et al, "A Simple Mechanism for Efficient Barrier Synchronization in MIMD Machines", Proc. 1990 ICPP, pp. 195-198.
• 5: L. Bomans and D. Roose, "Communications Benchmarks for the iPSC/2", Hypercube and Distributed Computers, Elsevier Science Publishers B.V., North-Holland, 1989.
• 6: T.H. Dunigan, "Hypercube Performance", Hypereube Multiprocessors 1987, SIAM, Philadelphia, PA, 1987.
• 7: S.S. Fried, "Personal Supercomputing with the Intei i860', BYTE, Jan. 1991, pp.347-358.
• 8: D.C. Grunwald and D.A Reed, "Benchmarking Hypercube Hardware and Software", Hypercube Multiprocessors 1987, SIAM, Philadelphia, Pa, 1987.
• 9: M.U Mathews, "Hypercube Software Performance Metrics', Hvpercube Multiprocessors 1987, SIAM, Phila., Pa, 1987.
• 10: C. McCreary and H. Gill, "Efficient Exploitation of Concurrency Using Graph Decomposition', Proc. 1990 ICPP, Vol. II, pp. 199-203.
• 11: W. Mendenhall and T. Sincich, A Second Course in Business Statistics: Regression Analysis, Dellen Publishing Company, San Francisco, 1989.
• 12: S. F. Nugent, The iPSC/2 direct-connect communications technology, Proceedings of the third conference on Hypercube concurrent computers and applications: Architecture, software, computer systems, and general issues, p.51-60, January 19-20, 1988, Pasadena, California, United States [doi>10.1145/62297.62305]
• 13: S. Nugent, Personal Communication, May 29, 1991.
• 14: Krishnan Padmanabhan, Cube structures for multiprocessors, Communications of the ACM, v.33 n.1, p.43-52, Jan. 1990 [doi>10.1145/76372.76376]
• 15: P. Pierce, The NX/2 operating system, Proceedings of the third conference on Hypercube concurrent computers and applications: Architecture, software, computer systems, and general issues, p.384-390, January 19-20, 1988, Pasadena, California, United States [doi>10.1145/62297.62341]
• 16: J. Ouinlan and D. Siewiorek, "Baseline measurements on the Mark II Hypercube", Hypercube Multiprocessors |987, SIAM, Philadelphia, Pa, 1987.
• 17: B. Stramm and F. Berman, "How Good is Good?", Tech Report CS90-169, UCSD, 1990.
• 18: D.W. Walker, et al, "A Comparison of the Performance of the Cal Tech Mark II Hypercube and the Elxsi 6400", Hypercube Multiprocessors 1987, SIAM, Philadelphia, Pa, 1987.
• 19: A. Zaafrani, H. Deitz, and M.T. O'Kcefe, "Static Scheduling for Barrier MIMD Architectures", Proc. 1990 ICPP, pp. 187-194.