Evaluation of buffer queries in spatial databases

Abstract

A class of commonly asked queries in a spatial database is known as buffer queries. An example of such a query is to “find housepower line pairs that are within 50 meters of each other.” A buffer query involves two spatial data sets and a distance d. The answer to this query are pairs of objects from the two input sets that are within distance d of each other. This paper addresses the problem of how to evaluate this class of queries efficiently. Geometric objects are used to denote the shape and location of spatial objects. Two objects are within distance d precisely when their minimum distance (minDist) is. A fundamental problem with buffer query evaluation is to find an effective algorithm for solving the minDist problem. Such an algorithm is found and its desirability is demonstrated. Finding a fast minDist algorithm is the first step to evaluate a buffer query efficiently. It is observed that many, and even most, candidates can be determined to be in the answer without resorting to the relatively expensive minDist operation. A candidate is first evaluated with the least expensive technique - called 0-object filtering. If it fails, a more costly operation, called 1-object filtering, is applied. Finally, if both filterings fail, the most expensive minDist algorithm is invoked. To show the effectiveness of these techniques, they are incorporated into the tree join algorithm and tested with real-life as well as synthetic data sets. Extensive experiments show that the proposed algorithm outperforms existing techniques by a wide margin in both the execution time as well as IO accesses. More importantly, the performance gain improves drastically with the increase of distance values.