Parallel processing of multicomponent seismic data
- PDF / 96,623 Bytes
- 5 Pages / 595.276 x 793.701 pts Page_size
- 11 Downloads / 194 Views
PARALLEL PROCESSING OF MULTICOMPONENT SEISMIC DATA UDC 681.3
V. V. Falfushinsky
Abstract. An algorithm for processing multicomponent seismic data is proposed. It is implemented in and its performance is measured on the Inparcom cluster. Several improvements are applied to speed up the program and to reduce the filesystem load, in particular, local folders are used to store temporary files of separate processes and indexing is used to speed up navigation through large files. Keywords: parallel programming, seismic data processing, optimization of cluster computing, large dataset. INTRODUCTION With the development of new technologies, the amount of results of investigations in seismic prospecting has increased. As is well known, many data of seismic observations are inexact and sometimes are distorted as a result of some disturbances (noises). The use of some conventional personal computer with a view to eliminating disturbances (noises) is inexpedient by virtue of its small computational capability. The execution of computations on modern multinuclear personal computers also does not provide any saving of time. Therefore, to proces such data, other parallel computing facilities are used among which the following systems are most efficient: • grid systems; • cloud computing systems; • cluster computing systems; • calculators with accelerators of graphics processing (GPUs). The problem of parallel processing of multicomponent seismic data is connected, first of all, with processing mass data. In view of an insufficient throughput of networks, the use of grid systems and cloud computing systems for parallelization is inexpedient. Calculators with GPUs can be applied, but an insufficient memory size of some types of such devices restricts their capability. To perform required calculations, the cluster complex Inparcom-256 was used, which was developed by the V. M. Glushkov Institute of Cybernetics together with the GNPP “Elektronmash.” PROCESSING OF SEISMIC PROSPECTING DATA The typical scheme of seismic investigation of the earth’s crust is as follows [1]. Seismic receivers (geophones) capable of registering their displacements in space are placed on the Earth’s surface along a planned seismic profile. A source of the seismic signal, namely, a definite number of vibration machines synchronously striking the ground with heavy hammers are placed along the same profile. As a result of action of this source, an acoustic wave is excited in the earth’s crust and is propagated downwards and laterally from the source at the velocity of sound. On the seismic boundary, i.e., on the sound speed drop boundary in the bedrock, wave refraction and wave reflection take place. The reflected wave comes back in the opposite direction, reaches the surface, and is registered by geophones. Each geophone writes (registers) V. M. Glushkov Institute of Cybernetics, National Academy of Sciences of Ukraine, Kyiv, Ukraine, [email protected]. Translated from Kibernetika i Sistemnyi Analiz, No. 2, pp. 181–186, March–April 2011. Orig
Data Loading...
