ABSTRACT
The LLGrid system of supercomputers allows hundreds of researchers simultaneous interactive access to large amounts of processing and storage for development and testing of their sensor processing algorithms. The requirements of the LLGrid user base are as diverse as the sensors they are developing: sonar, radar, infrared, optical, hyperspectral, video, bio, and cyber. LLGrid users are first and foremost scientists and engineers. The LLGrid system has been optimized to help the users get through the above steps with as little effort as possible. Benchmarking plays two important roles in LLGrid. The first is acceptance testing of the hardware to verify that the vendor has provided what was specified. The second is to optimize the software tools in the LLGrid software stack. LLGrid is a fleet of dozens of clusters, made of mostly commodity components. Distributed arrays is the dominant programming model on LLGrid.
