Title :
Parallel SAR processor using PVM on a Beowulf cluster
Author :
Inggs, M.R. ; Bennett, T.G.
Author_Institution :
Dept. of Electr. Eng., Cape Town Univ., Rondebosch, South Africa
Abstract :
Beowulf cluster computers (often called a "pile of PCs") have become common place, since they offer significant processing power at a fraction of the cost of custom, massively parallel machines. The processing needs of SAR naturally seem to point to a Beowulf cluster. However, as with all parallel computing, task granularity is a governing factor in the choice of compute node (processor, memory) and the interconnection (network technology and switch). This paper examines some of these issues in the design and testing of a SAR processor for a non real-time airborne radar. The system presented was constrained to fit to an existing, small Beowulf cluster
Keywords :
geophysical techniques; geophysics computing; parallel architectures; radar computing; remote sensing by radar; synthetic aperture radar; terrain mapping; Beowulf cluster; PVM; SAR; airborne radar; geophysical measurement technique; geophysics computing; land surface; parallel SAR processor; parallel computing; parallel processing; parallel virtual machine; pile of PCs; radar remote sensing; synthetic aperture radar; task granularity; terrain mapping; Computer networks; Concurrent computing; Costs; Multiprocessor interconnection networks; Parallel machines; Parallel processing; Personal communication networks; Switches; Synthetic aperture radar; Testing;
Conference_Titel :
Geoscience and Remote Sensing Symposium, 2001. IGARSS '01. IEEE 2001 International
Conference_Location :
Sydney, NSW
Print_ISBN :
0-7803-7031-7
DOI :
10.1109/IGARSS.2001.977093
