Improved parallel I/O via a two-phase run-time access strategy

ABSTRACT

As scientists expand their models to describe physical phenomena of increasingly large extent, I/O becomes crucial and a system with limited I/O capacity can severely constrain the performance of the entire program.We provide experimental results, performed on an lntel Touchtone Delta and nCUBE 2 I/O system, to show that the performance of existing parallel I/O systems can vary by several orders of magnitude as a function of the data access pattern of the parallel program. We then propose a two-phase access strategy, to be implemented in a runtime system, in which the data distribution on computational nodes is decoupled from storage distribution. Our experimental results show that performance improvements of several orders of magnitude over direct access based data distribution methods can be obtained, and that performance for most data access patterns can be improved to within a factor of 2 of the best performance. Further, the cost of redistribution is a very small fraction of the overall access cost.

REFERENCES

• 1: [DEBE91] E. DeBenedictis, P. Madams, "nCUBE's Parallel I/O with Unix Compatibility", Proceedings of the Sixth Distributed Memory Computing Conference, Portland, Oregon, April, 1991, pp. 270-277.
• 2: [DELR92] J.M. del Rosario, "High Performance Parallel I/O on the nCUBE 2", IEICE transactions, Japan, August, 1992.
• 3: James C. French , Terrence W. Pratt , Mriganka Das, Performance measurement of the concurrent file system of the Intel iPSC/2 hypercube, Journal of Parallel and Distributed Computing, v.17 n.1-2, p.115-121, Jan./Feb. 1993 [doi>10.1006/jpdc.1993.1009]
• 4: [HPCC91] "Grand Challenges: High Performance Computing and Communications", A Report by the Committee on Physical, Mathematical, and Engineering Sciences, Office of Science and Technology Policy, 1991.
• 5: [INTE92] Intel, "A Touchstone Delta System Description", Intel, Portland, Oregon, 1992. Intel Advanced Information.
• 6: [NCUB92] nCUBE, "nCUBE 2 Systems: Technical Overview", nCUBE, Foster City, California, 1992.
• 7: [PIER89] P. Pierce, "A concurrent file system for a highly parallel mass storage subsystem", Proceedings of the Fourth Conference on Hypercubes, Concurrent Computers, and Applications, pp. 161-6, 1989.

INDEX TERMS

Primary Classification:
  F. Theory of Computation
  F.1 COMPUTATION BY ABSTRACT DEVICES
      F.1.2 Modes of Computation
          Subjects: Parallelism and concurrency

Additional Classification:
  B. Hardware
  B.4 INPUT/OUTPUT AND DATA COMMUNICATIONS

  C. Computer Systems Organization

General Terms:
Design, Experimentation, Measurement, Performance, Theory

Keywords:
data distribution, operating system support, parallel I/O, parallel architecture, performance evaluation