A Parallel Nearest Neighbor Algorithm for Skyline Computation Using Map Reduce

Abstract

The Skyline is one of the most important operator in multi-criteria decision making and can be useful for many applications such as customer information services, decision support and decision making systems. It selects the best tuples from a multi-dimensional database. The main problem of using Skyline is that the queries that compute it can be computationally expensive, so the best solution is to do it using a parallelized approach. In this paper we propose a parallel algorithm based on nearest neighbor search to compute the skyline using the Map Reduce framework. This method consist on computing in a considered region the nearest neighbor point to the origin and partitions the region where each new region is obtained by adding the constraint that the coordinate with respect to a dimension is upper bounded by that of the computed point in the same dimension. The algorithm applies the same method recursively through these regions of the partition computed at each step. We provide a parallel approach for this method based on the Map Reduce framework to come up with a solution to this problem by handling the independent regions in parallel using mappers and reducers.