GREO is an attached database processor developed for midrange business computers. Since 1989, thousands of GREO’S have been shipped in the Japanese midrange market.

GREO is the first commercial product of a high. speed sorting hardware, or a hardwan sorter. GREO implements the O (n) time pipelined merge sort algorithm with several enhancements[1]. To fully utilize this sorting capability, the language processor for GREO compiles a given query into a sorting-oriented dataflow graph, which is in turn executed by the hardware sorter in conjunction with multi-microprocessors. Another unique aspect of GREO is that it is a commercial product developed in collaboration between a university and an industry(which are actually authors’ current affiliations). In spite of not a few “academic flavors” in its architecture, it succeeded in faithfully meeting practical customers requirements, ie, improving performance of a variety of existing(relational and even non-relational) business applications. The paper describes both of its technical and practical aspects. First, the query processing model underlying GREO architecture is presented. Its architectural features are then described. Our experiences on making research results an actual commercial product are also reported.