Output-optimal Parallel Algorithms for Similarity Joins
Xiao Hu, Yufei Tao, Ke Yi · 2017
Parallel join algorithms have received much attention in recent years, due to the rapid development of massively parallel systems such as MapReduce and Spark. In the database theory community, most efforts have been focused on studying worst-optimal algorithms. However, the worst-case optimality of these join algorithms relies on the hard instances having very large output sizes. In the case of a two-relation join, the hard instance is just a Cartesian product, with an output size that is quadratic in the input size.