A Fast Synthetic Aperture Radar Raw Data Simulation Using Cloud Computing.

Area of Science:

• Remote Sensing
• Computational Science
• Radar Engineering

Background:

• Synthetic Aperture Radar (SAR) raw data simulation is crucial for radar system design and imaging algorithms.
• Increasing data volumes and simulation times pose significant computational and I/O challenges.
• Existing high-performance computing (HPC) methods often struggle with I/O bottlenecks.

Purpose of the Study:

• To develop a cloud computing-based algorithm for accelerating SAR raw data simulation.
• To overcome the limitations of traditional HPC methods in handling large datasets.
• To improve both the computational speed and I/O efficiency of SAR raw data simulation.

Main Methods:

• Implementation of a cloud computing algorithm utilizing the MapReduce model for accelerated computation.
• Leveraging the Hadoop Distributed File System (HDFS) to address input/output (I/O) bottlenecks.
• Development of three optimization strategies focusing on programming models, HDFS configuration, and scheduling.

Main Results:

• Achieved a 4x speedup compared to the baseline serial approach in an 8-node cloud environment.
• Each optimization strategy individually improved performance by approximately 20%.
• Demonstrated the algorithm's effectiveness in solving computing-intensive and data-intensive issues in SAR simulation.

Conclusions:

• The proposed cloud computing algorithm effectively addresses the computational and data-intensive challenges in SAR raw data simulation.
• The MapReduce model offers advantages for irregular parallel accumulations in SAR simulation, outperforming GPU-based methods in certain aspects.
• The algorithm is scalable and extensible for large-scale computing, offering significant acceleration potential for radar research and development.