PARALLEL RELATIONAL OPERATIONS USING CLUSTERED SURROGATE FILES ON SHARED-NOTHING MULTIPROCESSORS

This paper presents data parallel algorithms for selection and join op erations based on clustered concatenated code word (CCW) surrogate fil es for very large relational databases. A multiattribute indexed exten dible hashing scheme is developed to cluster the surrogate files, and the performance of the parallel selection and join algorithms art. eva luated for a shared-nothing multiprocessor system. The proposed algori thms have an almost linear speedup as the number of processors increas es, and the proposed join algorithm requires less processing time comp ared to the parallel hybrid-hash join algorithm. (C) 1998 Elsevier Sci ence Inc. All rights reserved.