SEPT: an efficient skyline join algorithm on massive data

Abstract

Skyline join is an important operation in many applications to return all join tuples that are not dominated by any other join tuples. It is found that the existing algorithms cannot process skyline join on massive data efficiently. This paper presents a novel skyline join algorithm SEPT on massive data. SEPT utilizes sorted positional index lists with join information which require low space overhead to reduce I/O cost significantly. The sorted positional index list is constructed for each potential skyline attribute in the joined tables and is arranged in ascending order of the attribute. SEPT consists of two phases. In phase one, SEPT obtains candidate join positional index pairs of skyline join results. During retrieving the sorted positional index lists, SEPT performs pruning on candidate join positional index pairs in order to discard the candidates whose corresponding join tuples are not skyline join results. In phase two, SEPT exploits the obtained candidate join positional index pairs to get skyline join results by a selective and sequential scan on the tables. The experimental results on synthetic and real data sets show that SEPT has a significant advantage over the existing skyline join algorithms.