Open AccessIT SOLUTIONS FOR JINR TASKS ON THE “GOVORUN” SUPERCOMPUTER
Author(s) -
D. V. Podgainy,
D. Belyakov,
Andrey Nechaevskiy,
O. Streltsova,
А. А. Воронцов,
M. Zuev
Publication year - 2021
Publication title -
9th international conference "distributed computing and grid technologies in science and education"
Language(s) - English
Resource type - Conference proceedings
DOI - 10.54546/mlit.2021.95.45.001
Subject(s) - supercomputer , computer science , graphics , parallel computing , computational science , massively parallel , general purpose computing on graphics processing units , task (project management) , big data , cuda , volume (thermodynamics) , distributed computing , operating system , management , economics , physics , quantum mechanics
The “Govorun” supercomputer is a heterogeneous computing system that contains computingarchitectures of different types, including graphics accelerators. The given architecture of thesupercomputer allows users to choose optimal computing facilities for their tasks.To enhance the efficiency of solving user tasks, as well as to expand the efficiency of utilizing bothcomputing resources and data processing and storage resources, a number of special IT solutions havebeen implemented on the “Govorun” supercomputer. A hierarchical hyperconverged data processingand storage system with a software-defined architecture is referred to the first type of IT solutions. Theimplementation of this system is caused by the fact that modern supercomputers are used not only astraditional computing environments for carrying out massively parallel calculations, but also assystems for Big Data analysis and artificial intelligence tasks that arise in different scientific andapplied tasks. The second type of IT solutions lies in resource orchestration, which means thatcomputational elements (CPU cores and graphics accelerators) and data storage elements (SSDs) formindependent computing and data storage fields. Due to it, the user can allocate for his task the requirednumber and type of compute nodes (including the required number of graphics accelerators), as wellas the required volume and type of data storage systems.