Using Run-Time Uncertainty to Robustly Schedule Parallel Computation
by R.D. Dietz, T.L. Casavant, T.E. Scheetz, T.A. Braun, M.S. Andersiand
Abstract:
Increasingly, feedback of measured run-time information is being used in the optimization of computation execution. Because of this, the need for a model relating the static view of a computation to its runtime variance is becoming more important. Recently, we have described such a model which uses the notion of uncertainty to provide bounds on key scheduling parameters of the run-time computation. In this paper, we demonstrate how our model provides a foundation for robust parallel scheduling, i.e., scheduling that optimizes for computation execution in the presence of run-time variance. While this work was inspired by our previous study of uncertainty due to measurement intrusion, the scheduling paradigm presented here represents a broader, more general application of the uncertainty concept.
Keywords: theory
Source:
R.D. Dietz, T.L. Casavant, T.E. Scheetz, T.A. Braun, M.S. Andersiand, Using Run-Time Uncertainty to Robustly Schedule Parallel Computation. In V. Malyshkin (ed.),
Parallel Computing Technologies: Proceedings of the 4th International Conference,
Lect. Notes in Comp. Sci., Vol. 1277, Springer, 1997, pp. 13-24