Abstract:
Mathematical modeling allows to create a digital model of a product, the so-called digital twin. Such digital twins are quite complex mathematical models, the description of which, as well as the calculation of their parameters and characteristics, is a very serious computational problem that can be solved on high-performance computing systems. However, a building a high-performance supercomputing system designed for solving problems of computer simulation of various physical processes is a difficult time-consuming task. The paper discusses basic approaches to design a login nodes subsystem for such high-performance supercomputing system, allowing to systematize and simplify the process of its development. The analysis of the main factors affecting the structure and composition of the login nodes subsystem is carried out. These factors include the number of potential users of a high-performance computing system; availability of categories of users of the system and their percentage; characteristics of computing tasks solved on the system; hardware platform chosen for building systems. An example of a methodology for calculating of the login nodes subsystem size is given.
Citation:
S. Yu. Mokshin, “General approaches to the design of the access subsystem of high-performance computing systems”, Proceedings of ISP RAS, 32:4 (2020), 41–52
\Bibitem{Mok20}
\by S.~Yu.~Mokshin
\paper General approaches to the design of the access subsystem of high-performance computing systems
\jour Proceedings of ISP RAS
\yr 2020
\vol 32
\issue 4
\pages 41--52
\mathnet{http://mi.mathnet.ru/tisp523}
\crossref{https://doi.org/10.15514/ISPRAS-2020-32(4)-3}
Linking options:
https://www.mathnet.ru/eng/tisp523
https://www.mathnet.ru/eng/tisp/v32/i4/p41
This publication is cited in the following 1 articles: