Most database queries can be reduced to a few distinct classes of requests. These are (1) finding the k/sup th/ smallest element in an unordered set S, (2) finding the k/sup th/ through the (k+i)/sup th/ smallest elements in an unordered set S, and (3) return the record with a specific key. This paper presents a fast, efficient selection based system implemented on the DAP 510 for each of these query types. The DAP or the distributed array of processors is a commercially available array processor, a product of Active Memory Technology. This massively parallel computer attaches to a host computer as a peripheral processor. It differs from the conventional serial processor in that it can perform the same operation on many items of data in parallel. >