Processing partially specified queries over high-dimensional databases

摘要

In recent years, high-dimensional database applications deal with multidimensional ad hoc queries that refer to an arbitrary number of arbitrarily (unpredictably) chosen dimensions of high-dimensional data. This paper thoroughly and systematically investigates possible secondary-storage based solutions to the problems of processing multidimensional ad hoc query in transactional or semi-transactional environments. Then a complementary solution called the indexed and transposed access method (ITAM) is proposed. This method is based on two complementary measures – a multidimensional access method and a proposed access method called the Opus-path. The performance of multidimensional access methods deteriorates rapidly as the ratio of query dimensionality to data dimensionality decreases. On the other hand, the Opus-path shows retrieval performance that is actually better when the ratio is low.