Optimizing Multiple Top-K Queries over Joins

Advanced Data Mining applications require more and more support from relational database engines. Especially clustering applications in high dimensional features space demand a proper support of multiple Top-k queries in order to perform projected clustering. Although some research tackles to problem of optimizing restricted ranking (top-k) queries, there is no solution considering more than one single ranking criterion. This deficit optimizing multiple Topk queries over joins is targeted by this paper from two perspectives. On the one hand, we propose a minimal but quite handy extension of SQL to express multiple top-k queries. On the other hand, we propose an optimized hash join strategy to efficiently execute this type of queries. Extensive experiments conducted in this context show the feasibility of our proposal.