Research On Trajectory Big Data Management Upon HBase

With the rapid development of mobile internet and positioning technology,mobile devices with location-awareness are becoming more popular.Nowadays,people can widely obtain trajectory data of various mobile objects and form trajectory big data.In order to fully tap the value of trajectory big data,effective management of trajectory big data should be carried out first.In recent years,there have been some studies on trajectory big data management based on big data frameworks such as Hadoop,HBase,and Spark.However,many studies store trajectories as points or segments,which cannot effectively support analysis based on complete trajectories.At the same time,most studies have oversimplified the handling of trajectory time attributes,resulting in difficulty in quickly filtering invalid records through time constraints in queries.In addition,existing studies only consider overlapping or containment range queries,which cannot efficiently meet the diverse needs of complex trajectory queries.To address these issues,this paper first studies the dimensionality reduction coding of trajectory spatiotemporal attributes and its spatiotemporal filtering capability.Then,based on the HBase indexing mechanism,XZST and IDXZT indexes are designed to support various trajectory queries.Finally,combined with the characteristics of HBase,a prototype system for trajectory big data management is built.The main work is as follows:(1)The dimensionality reduction coding of trajectory spatiotemporal attributes and its spatiotemporal filtering strategy are designed and implemented.In this paper,XZ and XZT space filling curves are used to encode the spatial sequence and time range of trajectories respectively,so that high-dimensional objects can be uniquely mapped to the real domain while preserving their respective proximity relationships in highdimensional space.Then,based on the characteristics of the two types of encoding,a filtering strategy for overlapping and containment range in spatial and temporal dimensions based on encoding is proposed.(2)Indexes for trajectory data named XZST and IDXZT are designed,and multiple trajectory query types are implemented.In this paper,trajectories are managed in complete,and the trajectory spatiotemporal encoding and related attribute information are concatenated in a certain order to form XZST and IDXZT indexes,which are indexed as row key in HBase.Then,spatiotemporal range queries and mobile object queries that can be independently configured for spatiotemporal constraints are designed.Finally,based on the spatiotemporal range query,KNN query is designed.(3)A prototype system for trajectory big data management based on HBase is designed and implemented.The data storage layer is responsible for data storage and import,the encoding and indexing layer implements the trajectory encoding and indexing methods,and the trajectory query layer implements multiple query functions,and implements server-side filtering and secondary indexing queries based on HBase coprocessors.The experimental results based on a real dataset show that the encodinglevel filtering strategy and server-side filtering proposed in this paper can improve the efficiency of trajectory queries,and the auxiliary index only requires a small amount of additional index storage to avoid full table scanning,which can also enrich the index query strategy of trajectory data.