C. Layer and H.-J. Pfleiderer (Germany)
Parallel Computing, Data Mining, Associative Access, Sorting Algorithms, Vertical Counting.
Relying on an associative access method for the retrieval of information in database searching systems, this paper presents a novel architecture for the parallel processing of bit sliced signature files. It describes a new scheme for the construction of a search algorithm addressing associative memories and thus reducing the number of data accesses. The retrieval procedure is based on approximate matching of text strings for which a variety of bit vector text signa tures are proven from previous studies [4, 6, 8] to be appro priate for the implementation of parallel database systems at the hardware level. Therefore, efforts have been made in the development of vertical sorting algorithms handling data bit by bit in parallel, directly in the memory. Com bined with a vertical storage functionality and a Hamming distance calculation, the vertical sorting method can signifi cantly improve the overall processing time and the memory usage of the complete database system.
Important Links:
Go Back
