Efficient parallel top-k computation algorithm using symmetry breaking

Abstract

A key problem of relational database is to aggregate different values of the same object and find the first k objects with highest overall values. Many sequential algorithms have been proposed to solve this problem. In this paper, we propose a new parallel algorithm using symmetry breaking strategy. New algorithm is proved to be instance optimal. Experiment results on both synthetic and real data also show that new algorithm costs fewer accesses, compared to previous algorithms. © 2010 IEEE.