Block-Join: A Partition-Based Method for Processing Spatio-Temporal Joins

Abstract

Spatio-temporal joins are important operations in spatio-temporal databases. The rapid increase in the amount of spatio-temporal objects makes the cost of spatio-temporal joins expensive and requires an efficient method for spatio-temporal joins. In this paper, we propose a block-join method for spatio-temporal joins by partitioned blocks. We first partition the entire spatio-temporal data space of two trajectory datasets into equal-sized blocks. Spatio-temporal objects with similar spatio-temporal attributes will be split into the same block. To achieve a uniform distribution of trajectories inside a block called block trajectories, we merge two blocks into one block called unequal-sized blocks. Then, we evaluate block trajectories in the same and adjacent blocks to get pairs satisfying the spatio-temporal join conditions. The pairs are sorted and removed duplicated pairs to get precise results. Using both real and synthetic datasets, we carry out comprehensive experiments in a prototype database system to evaluate the efficiency of our methods. The experimental results show that our approach outperforms alternative methods in the system by a factor of 2-10x on large datasets.