No abstract

An associative processor called RAP has been designed to provide hardware support for the use and manipulation of databases. RAP is particularly suited for supporting relational databases. In this paper, the relational operations provided by the RAP hardware are described, and a representative approach to providing the same relational operations with conventional software and hardware is devised. Analytic models are constructed for RAP and the conventional system. The execution times of several of the operations are shown to be vastly improved with RAP for large relations.

xiic xteiaLxuiiH.! ^.ssucictLive rrocessor (n.tt..r / / is an experimental "backend" cellular processor for implementing data base management systems. RAP is particularly well suited to supporting Codd's relational model of data. The capacity of a RAP device implemented with current IC and memory technology can be estimated to contain 10 8 to 10 9 bits of associativeiy processable data. Because many data bases are larger, a virtual memory environment for RAP has been proposed and its performance simulated. The environment incorporates conventional memories for bulk storage and a single RAP processor-both controlled by a general purpose ±ront-enu computer, me system requires that the entire relational data base be divided into pages of size equal to one RAP cell memory. A buffer memory is added to RAP to permit the overlap of paging with processing. It has been found that user environments containing small relations or queries exhibiting either long processing times relative to paging requirements or some "locality" (defined as the degree to which sequences of queries reference some relations more than others) can efficiently page data between large data bases and data base machines without significant losses in performance.

No abstract

No abstract