Search academic texts

Information × Registration Number 0212U004747, 0111U004833 , R & D reports Title Application of the GRID-technology to the astrophysical computing intensive tasks and upgrade/development of the GRID cluster of MAO NANU popup.stage_title Head Berczik P.P., Registration Date 25-01-2012 Organization Main Astronomical Observatory popup.description2 Construction of new and strengthening of existent clusters and grid- platform of access in leading academic establishments 4.1. The use of grid- technology is in fundamental scientific and scientifically-applied researches. keywords: Hybrid cluster technologies, grid-infrastructure, GPGPU of calculation, CUDA, OpenCL. Processing :model-independent results that is got within the framework of cosmological designs will be drawn on for determination of nature of particles-candidates on the role of dark matter and dark energy. Among the actual tasks of forming and evolution of large-scale structures of Universe, that require calculations from by the use grid- technology and creation of corresponding software, such are executed: a) Forming and complex physical and chemical evolution of galaxies and accumulations of galaxies; b) Dynamics and evolution of confluence of galaxies; с) Dynamic and spatial descriptions of the isolated galaxies in comparing to the galaxies that are in dense. It is worked out and set on a grid-site of MAO NAN of Ukraine runtime environment, that allows to start by means of grid of GPU and CPU of version of programmatic package for the direct design of task of many bodies of phi - GRAPE. Testing of programmatic package of phi is executed - GRAPE in a grid environment. At a calculation with the use of GPU there is the considerable falling of speed on the less values of N. But even in this case speed of calculations with the use of GPU in 10 - 200 times higher, than with the use of CPU. On the new test systems with the use of two different graphic accelerating of GF 570 GTX and ATI HD 6970 the comparative astrophysical calculations of dynamic design of task of N were conducted - body with the use of the newest technology of OpenCL. Our calculations convincingly testify to the high yield of both graphic accelerating (about 150 Gflops on one map) in comparing to multinuclear CPU on both test knots. Table of contents of project in part of the job of a 1.4 program processing : modernisation is executed for the increase of efficiency of work of grid- cluster of MAO NAN of Ukraine. a а) cluster is extended by 2 knots that equipped two GPU by the calculators of AMD Radeon, and two calculators of NVidia of б) are extend a disk array on 6Тб and в) the existent knots of cluster it is reconditioned modern graphic calculators. Modernisation power of cluster allowed to promote to ~2TGlops at the decision of tasks of N-body facilities of GPGPU (CUDA and OpenCL). Product Description popup.authors Берцик Петер Петерович Велесь Олександр Анатолійович Зінченко І.А. popup.nrat_date 2020-04-02 Close