The integration of HPC systems in a test and evaluation real-time environment

In simulated altitude testing of turbine engines, large volumes of data - on the or-tiel of 500, 000 samples/sec -- ale acquired, processed validated, and analyzed. Validation of these data is a labor-intensive effort. Mathematical models have been developed to improve the speed and effectiveness of validation, and the need to run these models in I cal time has been noted, Current AEDC acquisition systems use reflective memory to distribute data among UNIX and NT systems, where they are distributed in a client-server fashion for display and processing. One of these UNIX workstations, The Data Distribution Server, a UNIX workstation, sends the data over a Fast Ethernet link to the High Performance Computer (HPC) system, where the data validation code is run. This link uses of existing fiber-optic cable plant, inexpensive Fast Ethernet interfaces that are now standard on most systems, and commodity networking gear to provide a cost-effective capability to supply a near real-time data stream to the data validation code running on the HPC system. This system reduces the cost of turbine engine testing by reducing the time required to validate data and the number of test points that must be repeated because of bad data. This paper describes ou, approach for transmitting huge quantities of real-time data to a remote HPC system by capitalizing on existing infrastructure.