Abstract
We discuss in the paper the influence exerted by applied mathematics techniques on the progress of the architecture and performance of computing systems. We focus on computational algorithms aimed at solving the problem of fault tolerance which is topical for future exaFLOPS and super-exaFLOPS systems.
This work was supported by the Russian Foundation for Basic Research (project No. 17-07-01604-a).
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Principle of “geometric parallelism”, also referred to as domain decomposition, is a parallelization method widely used in solving problems of mathematical physics. As specified by this technique, the computational domain is divided into subdomains, then each processor executes the same code on its own individual subdomain and exchanges boundary data with neighboring processors.
References
Davydov, A.A., Lacis, A.O., Lutsky, A.E., Smoliyanov, Yu.P., Chetverushkin, B.N., Shilnikov, E.V.: The hybrid supercomputer MVS-express. Doklady Math. 82(2), 816–819 (2017). https://doi.org/10.1134/S1064562410050364
Gibridnyi vychislitel’nyi klaster k-100 [hybrid computer cluster k-100]. www.kiam.ru/MVS/resourses/k100.html
Gorobets, A.V., Sukov, S.A., Zhelezniakov, A.O., Bogdanov, P.B., Chetverushkin, B.N.: Rasshirenie dvukhurovnevogo rasparallelivaniia MPI+openmp posredstvom opencl dlia gazodinamicheskikh raschetov na geterogennykh sistemakh [extension of two-level parallelization MPI+openmp through opencl for hydrodynamic computations on heterogeneous systems]. Vestnik Iuzhno-Ural’skogo gosudarstvennogo universiteta 9, 76–86 (2011)
Bland, W.: Int. J. High Perform. Comput. Appl. 27(3), 244–254 (2013)
Cappello, F.: Int. J. High Perform. Comput. Appl. 23(3), 212–226 (2009)
Cappello, F., Geist, A., Gropp, W., Kale, S., Kramer, B., Snir, M.: Int. J. High Perform. Comput. Appl. 1(1), 1–28 (2014)
Snir, M., et al.: Int. J. High Perform. Comput. Appl. 28(2), 129–173 (2014)
Chetverushkin, B.N.: Kineticheskie skhemy i kvazigazodinamicheskaia sistema uravnenii [Kinetic schemes and quasi-gasdynamic system of equations]. Max Press, Moscow (2004)
Chetverushkin, B.N.: Resolution limits of continuous media mode and their mathematical formulations. Math. Models Comput. Simul. 5, 266–279 (2013). https://doi.org/10.1134/S2070048213030034
Sedov, L.I.: Metody podobiia i razmernosti v mekhanike. [Similarity and dimensionality methods in mechanics]. Nauka, Moscow (1977)
Zlotnik, A.A., Chetverushkin, B.N.: Entropy balance for theone-dimensional hyperbolic quasi-gasdynamic system of equations. Doklady Math. 95(3), 276–281 (2017). https://doi.org/10.1134/S106456241703005X
Chetverushkin, B.N., Zlotnik, A.A.: On some properties of multidimensional hyperbolic quasi-gasdynamic systems of equations. Russ. J. Math. Phys. 24, 299–309 (2017). https://doi.org/10.1134/S1061920817030037
D’Aschenzo, N., Chetverushkin, B.N., Saveliev, V.I.: On an algorithm for solving parabolic and elliptic equations. Comp. Math. Math. Phys. 55(8), 1290–1297 (2015). https://doi.org/10.1134/S0965542515080035
Chetverushkin, B., D’Ascenzo, N., Saveliev, A., Saveliev, V.: A kinetic model for magnetogasdynamics. Math. Mod. Comp. Simul. 9(5), 544–553 (2017). https://doi.org/10.1134/S2070048217050039
Zlotnik, A.A., Chetverushkin, B.N.: O parabolichnosti kvazigazo-dinamicheskoi sistemy uravnenii, ee giperbolicheskoi 2-go poriadka modifikatsii i ustoichivosti malykh vozmushchenii dlia nikh [on parabolicity of quasi-gasdynamic system of equations, its 2nd order hyperbolic modification, and stability of small perturbations for them]. Zhurnal vychislitel’noi matematiki i matematicheskoi fiziki, pp. 445–472 (2008)
Chetverushkin, B.N., Yakobovskiy, M.V.: Vychislitel’nye algoritmy i otkazoustoichivost’ giperekzaflopsnykh vychislitel’nykh sistem [computational algorithms and fault-tolerance of hyper-exaflops computing systems]. Doklady Akademii nauk [Proc. Russ. Acad. Sci.] 472(1), 1–5 (2017). https://doi.org/10.7868/S0869565217010042
Bondarenko, A.A., Yakobovskiy, M.V.: Modelirovanie otkazov v vysoko-proizvoditel’nykh vychislitel’nykh sistemakh v ramkakh standarta mpi i ego rasshireniia ulfm. [simulation of failures in high-performance computing systems within the mpi standard and its ulfm extension]. Vestnik Iuzhno-Ural’skogo gosudarstvennogo universiteta 4(3), 5–12 (2015). www.mathnet.ru/links/76f8ebac383f6603304d157a36123407/vyurv1.pdf
Bondarenko, A.A., Kornilina, M.A., Yakobovskiy, M.V.: Fault tolerant algorithm for HPC. In: Supercomputing in Scientific and Industrial Problems KIAM RAS, p. 8 (2016). www.kiam.ru/SSIP/SSIP_abstracts.pdf#page=9
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2019 Springer Nature Switzerland AG
About this paper
Cite this paper
Chetverushkin, B.N., Yakobovskiy, M.V., Kornilina, M.A., Semenova, A.V. (2019). Numerical Algorithms for HPC Systems and Fault Tolerance. In: Sokolinsky, L., Zymbler, M. (eds) Parallel Computational Technologies. PCT 2019. Communications in Computer and Information Science, vol 1063. Springer, Cham. https://doi.org/10.1007/978-3-030-28163-2_3
Download citation
DOI: https://doi.org/10.1007/978-3-030-28163-2_3
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-28162-5
Online ISBN: 978-3-030-28163-2
eBook Packages: Computer ScienceComputer Science (R0)