Parallel synthetic aperture radar processing on workstation networks

Synthetic aperture radar (SAR) signal processing poses a significant challenge due to its very large computation and data storage requirements. This paper presents the computational requirements of a typical high resolution satellite SAR data processing scenario. A classification of approaches to partitioning the SAR problem for parallel processing is given. The suitability of networks of workstations (NOW) for SAR processing is analyzed for a number of partitioning approaches. The network throughputs required to support SAR processing on NOW are derived. SAR processing is found to demand extremely high network throughput that is difficult to achieve with today's technology.