The Design And Application Of Location Services Data Index In Map/Reduce Mode

Widely used for location-based services, provides the basis of the large number ofhuman trajectory data for research. Daily trajectory data of location services systemscan achieve TB or even PB. At the same time the diversity of location servicesbusiness applications result in inconsistent data format (such as GPS data, phone, data,etc.). So multi-dimensional retrieval for location services data needs to overcome thedifficulty is not only the amount of data, fast response time, but also a semi-structuredbusiness data.We formalize the definition of the existing location services data and six kinds ofretrieval methods. First we formalize the definition of location services data, andselected a point-to-trajectory space measurement standards based on it. Then wepropose the six retrieval methods.By contrasting to the existing trajectory indexing methods, we propose of theMulti_Layer temporal and spatial index algorithm in the Map/Reduce mode for lowtrajectory recognition. At the same time we descript the Distributed definition andoperation of a multi-dimensional indexing mechanism from the definition of thestructure of the index, the storage mode, as well as atomic operations, which detailsthe specific index build and retrieval operations.Combined indexing mechanism of Hbase and Multi_Layer algorithm, wepropose a Map/Reduce mode multidimensional retrieval framework. Through thecomparison of the experiment, the newly created multi-dimensional indexframework of the three dimensions for the TB/PB-level location services data(latitude and longitude range, time period, vehicle ID) retrieval speed can reachsecond response.We propose multi-dimensional indexing methods for location services data underMap/Reduce mode in order to achieve real-time query for TB/PB-level locationservices data, which include: the definition and operation of distributed index; how toapply the index to location services data query; application examples and comparativeexperiments. But the time dimension and spatial dimensions combined algorithm haveno further discussion in index structures based the Hadoop retrieval framework. Sothere is room for improvement in the response time. In addition, there is a test for the comparison of GPS trajectory data for two months due to the time, and more types oflocation services data can not be tested.