FACET: PC-parallel architecture for cost-efficient telemetry processing

NASA's Mission to Planet Earth (MTPE) is planning to launch the Earth Observing System (EOS) starting in 1998. The large number of planned remote sensing satellites will bring 500 Gigabytes of information per day. The EOS Data and Information System (EOSDIS) is responsible for ingesting and archiving this data. One important component of the EOSDIS system is the data operation, which involves extracting the packets and reconstructing and archiving the original remotely sensed data products. Due to transmission errors, the way data is sampled from the different sensors encoded, packets typically arrive out of order and perhaps with some of them missing or repeated. Many special hardware solutions have been proposed to solve this real-time problem. In this paper, we demonstrate a commercial off the shelf (COTS) solution. The hardware capitalizes on the progress made in the area of network of workstations (NOW), particularly PC-clusters. The software and algorithm exploit the data characteristics and parallelism in the telemetry stream to make use of load balancing and efficient parallel processing. It is shown that this solution can provide high-performance to cost and programmability.