Prepartitioning in MapReduce Processing of Group Nearest-Neighbor Query

Algorithms for processing closest-pairs and nearest-neighbors queries on big spatial data in parallel and distributed frameworks

“…In [55] we presented a modified version of the algorithm which utilizes a prepartitioning The experiments showed a great performance increase, compared to the previous version.…”

“…The algorithm presentation in this section is a unified approach of the ones presented in [52] and [53] and their differences will be noted when necessary. The algorithm modification in [55] will be presented separately, because it is quite different. We start with two spatial datasets of points, Q or Query and P or Training.…”

“…These experiments were presented in [55]. The major difference compared to [53] is the prepartitioning phase which preceeds the others and the prepartitioned dataset that effectively replaces the Training as input to all phases.…”

“…Finally, in [55] we presented a modified version of the algorithm in which there was a preliminary phase that prepartitioned the Training dataset. This proved to be very effective because there were several repeating point location functions scattered around the code and performed the same actions multiple times.…”

“…Later in [53] an improved version of the algorithm was presented, along with a SpatialHadoop implementation and a study of what happens internally during the operation of the algorithm, using specialized cluster wide metrics. In [55] a prepartitioning method was introduced for the Training, which proved to be very effective. All these versions of the algorithm were extensively tested using a large variety of configuration parameters and spatial datasets.…”