Processing high-speed stereo particle image velocimetry data with an integrated cluster supercomputer

This paper presents the design of a cluster supercomputer integrated with a Particle Image Velocimetry measurement system. The system was designed so that it is fairly portable and can fit within the power and cooling constrains of a simple experimental fluid dynamics lab. Several low power processor choices are compared and one node of the proposed systems is tested for its suitability as a compute node in such a cluster computer. Data management issues are discussed as well since such a system is able to process and store large amounts of data. By extending an existing Computational Fluid Dynamics standard to experimental fluid dynamics data several data management problems can be overcome and the data exchange between computations and experiments can be facilitated.