Application of SCIT supercomputers to develop and execute parallel geophysical programs
- PDF / 926,233 Bytes
- 14 Pages / 595.276 x 793.701 pts Page_size
- 32 Downloads / 130 Views
APPLICATION OF SCIT SUPERCOMPUTERS TO DEVELOP AND EXECUTE PARALLEL GEOPHYSICAL PROGRAMS V. G. Tulchinskya† and P. G. Tulchinskya‡
UDC 681.3
The experience of developing parallel programs for seismic exploration is described. Seismic datasets are usually extensive and have easily decomposable structure. These features result in similar paralleling techniques for different problems such as trace-by-trace processing, migrations, and finite-difference modeling. The paralleling project development is described in terms of a pattern language for parallel programming and Saaty’s analytic hierarchy process. The parallel data processing is analyzed for efficiency. Keywords: parallel programming, seismic data processing, optimization of cluster computing, large data sets. INTRODUCTION The problematics of parallel programming is well developed as far back as in 1970–1980. The recent outbreak of parallel computing is due to the continuous accumulation of data and increase in task size along with technological limitations on further increase of the CPU clock rate. At the hardware level, this tendency was manifested as a changeover to multicore computers, expansion of multicomputer clusters, emergence of parallel computers based on videocards. Indirectly, it led to a jump in the performance of LANs and disk storages. At the systemic level, the concepts of high performance computing, HPC, grid, and cloud computing indirectly contribute to open source programming and promotion of Linux, despite the high quality software for Windows clusters. Ukraine has not stood aloof from these developments due to the combination of scientists’ initiatives and the financial capacity of the state during economic growth. In recent years, the Ukrainian SCIT cluster has been ranked high in performance [1] and extensively used by remote users and several grid systems. A large number of smaller clusters were purchased from commercial suppliers or built by Ukrainian enterprises for their own needs. Because of the economic recession, significant investment in computer technology is unlikely in the near future. Therefore, there is a need for higher performance and efficient use of existing computers. One of the application domains for supercomputers is seismic data processing. 1. SEISMIC DATA PROCESSING Let us present a typical pattern of the seismic study of the Earth’s crust [2]. Receivers (geophones) capable to register their spatial movements are placed on the surface along the planned seismic profile. A seismic signal source (a buried charge or a group of vibrating machines that simultaneously strike the ground with heavy hammers) is placed along the same profile. An acoustic wave excited by the source in the Earth’s crust propagates with the sonic velocity downward and outward from the source. At the seismic boundary, i.e., boundary of the sonic velocity difference in rock, a refraction and a V. M. Glushkov Institute of Cybernetics, National Academy of Sciences of Ukraine, Kyiv, Ukraine, [email protected]; ‡[email protected]. Translated from Kiberne
Data Loading...
