PPAM 2013: Parallel Processing and Applied Mathematics pp 47-56 | Cite as
Adaptive Load Balancing for Massively Parallel Multi-Level Monte Carlo Solvers
Abstract
The Multi-Level Monte Carlo algorithm was shown to be a robust solver for uncertainty quantification in the solutions of multi-dimensional systems of stochastic conservation laws. For random fluxes or random initial data with large variances, the time step of the explicit time stepping scheme becomes random due to the random CFL stability restriction. Such sample path dependent complexity of the underlying deterministic solver renders our static load balancing of the MLMC algorithm very inefficient. We introduce an adaptive load balancing procedure based on two key ingredients: (1) pre-computation of the time step size for each draw of random inputs (realization), (2) distribution of the samples using the greedy algorithm to “workers” with heterogeneous speeds of execution. Numerical experiments showing strong scaling are presented.
Keywords
Uncertainty quantification Conservation laws Multi-level Monte Carlo FVM Load balancing Greedy algorithms Linear scalingNotes
Acknowledgments
This work is performed under ETH interdisciplinary research grant CH1-03 10-1 and CSCS production project grant ID S366.
References
- 1.ALSVID-UQ, v3.0. http://www.sam.math.ethz.ch/alsvid-uq
- 2.Dandamudi, S.P.: Sensitivity evaluation of dynamic load sharing in distributed systems. IEEE Concurrency 6(3), 62–72 (1998)CrossRefGoogle Scholar
- 3.Graham, R.L.: Bounds on multiprocessing timing anomalies. SIAM J. Appl. Math. 17(2), 416–429 (1969)CrossRefMATHMathSciNetGoogle Scholar
- 4.LeVeque, R.A.: Numerical Solution of Hyperbolic Conservation Laws. Cambridge University Press, Cambridge (2005)Google Scholar
- 5.Mishra, S., Risebro, N.H., Schwab, C., Tokareva, S.: Multi-level Monte Carlo finite volume methods for scalar conservation laws with random flux. Technical Report, SAM, 2012. http://www.sam.math.ethz.ch/reports/2012/35
- 6.Mishra, S., Schwab, C., Šukys, J.: Multi-level monte carlo finite volume methods for uncertainty quantification in nonlinear systems of balance laws. In: Bijl, H., Lucor, D., Mishra, S., Schwab, C. (eds.) Uncertainty Quantification in Computational Fluid Dynamics. Lecture Notes in Computational Science and Engineering, vol. 92, pp. 225–294. Springer, Zurich (2013)CrossRefGoogle Scholar
- 7.Mishra, S., Schwab, C.: Sparse tensor multi-level monte carlo finite volume methods for hyperbolic conservation laws with random initial data. Math. Comp. 280(81), 1979–2018 (2012)CrossRefMathSciNetGoogle Scholar
- 8.Šukys, J., Schwab, C., Mishra, S.: Multi-level Monte Carlo finite difference and finite volume methods for stochastic linear hyperbolic systems. In: Dick, J., et al. (eds.) Monte Carlo and Quasi-Monte Carlo Methods 2012, vol. 65, pp. 649–666. Springer, Berlin (2013) http://www.sam.math.ethz.ch/reports/2012/19
- 9.Šukys, J., Mishra, S., Schwab, C.: Static load balancing for multi-level Monte Carlo finite volume solvers. In: Wyrzykowski, R., Dongarra, J., Karczewski, K., Waśniewski, J. (eds.) PPAM 2011, Part I. LNCS, vol. 7203, pp. 245–254. Springer, Heidelberg (2012)Google Scholar
- 10.MPI: A Message-Passing Interface Standard. Version 2.2, 2009. http://www.mpi-forum.org/docs/mpi-2.2/mpi22-report.pdf
- 11.Rosa, Swiss National Supercomputing Center (CSCS), Lugano, www.cscs.ch