Field Programmable Gate Array (FPGA) technology is finding increasing use in the kinds of large-scale data acquisition tasks characteristic of high-energy physics, and it can also be used for specialized data processing. Graphics Processing Units (GPUs) containing large numbers of processor cores are finding increasing use in processing large arrays of data. We propose to combine these technologies in a Streaming Data Reduction Appliance for real-time analysis of experimental data. With a flexible set of streaming interfaces that can support either standards like Camera Link or more specialized interfaces, and a standard gigabit Ethernet back end interface, such a device could serve as a flexible component for a scalable acquisition and processing system for live data. We look forward to producing this device along with an open design form of documentation that makes the technology accessible to any scientist with a problem to solve. As test cases for the applicability of this technology, we target the DOE sponsored Dark Energy Survey (DES) and Large Scale Synoptic Telescope (LSST) projects. We also believe this technology has much wider applicability
