Implementing a high-rate, low-cost data processing system
G.T. Butulis, Curt Tilmes, J.K. Hord, M.J.C. Bur, A. J. Fleig · 2002
Large-scale processing systems for producing information from high data rate instruments have traditionally been built using large, expensive multi-processor computer systems. In the last few years, improvements in commodity computing systems and storage devices have brought them to the performance territory once held exclusively by the likes of DEC, SGI, Sun, and IBM. We are now building a data processing system using PC-class machines for the Ozone Monitoring Instrument (OMI) that will be launched in 2003. OMI has a 1Mbit/second data rate and will run continuously for five years. The OMI production system, which is based on the 3/sup rd/ generation Moderate Resolution Imaging Spectroradiometer (MODIS) data processing system, will use a heterogeneous set of low-cost processors, high-volume disks, and fast networks to provide impressive throughput for an exceptionally low cost. This approach provides an attractive alternative to the high initial outlay and continuing support costs of the traditional "heavy iron" systems.