A Graph Based Bi-level Index for Spatio-temporal Data Analysis with MapReduce

The boosting deployment of GPS devices in urban vehicles is leading to the collection of large volumes of GPS. Such massive spatial-temporal datasets challenges the efficiency and scalability of the query process during data analysis. In this paper, we introduce the MapReduce framework into the GPS data analysis system. Particularly, we built a graph based bi-level index to accelerate the spatial query processing. The key idea is that we use topological graph instead of traditional R-tree index to lock the space scope of the GPS data, due to the fact that vehicles are moving along the road network. This index is also packed by the PGP(Parallel graph packing)algorithm to ensure the scalability. Experimental results show that the speedup and scale up of our work are very efficient.