Optimization Of Space Time Trajectory Load Balancing Based On HBase

Today,with the rapid development of communication technology,a large amount of trajectory data is generated every day.Based on these rich trajectory data,many popular research directions have been derived.However,in the face of the increasingly complex track data mining at the application level,the requirements for the reading and writing performance of the underlying track data are also increasingly high.The traditional single-node storage scheme cannot cope with the growing demand for track point data set storage.In order to improve storage performance and capacity,distributed storage has been paid more attention.However,the existing spatiotemporal storage solutions are not optimistic in terms of range query performance.The main difficulties lie in the design of index construction schemes and the requirements for load balancing.Although distributed storage can significantly improve the write performance and expand the upper limit of storage capacity horizontally,the performance of spatial-temporal query is weaker than that of single-node storage based on its own limitations.At the same time,trajectory data has its particularity,and data is extremely prone to skewing.Although trajectory can be discretized,it will break the relationship between trajectory points,affect overall query performance,and further highlight the drawbacks of distributed processing.To solve the above problems,this paper proposes a distributed track storage framework(JUST+)running on HBase cluster.This framework reduces the dimension of the track based on the space filling curve,speeds up the index construction speed and improves the track writing performance.At the same time,the trajectory pre-zoning and load balancing strategy are designed to avoid the load tilt caused by the excessive concentration of trajectory points.The main contributions of this paper are as follows:(1)Spatiotemporal trajectory index construction based on spatial filling curve:Based on the scenario of bulk commodity logistics,this paper combs the trajectory characteristics of freight vehicles,analyzes the problems existing in traditional trajectory index construction schemes,and constructs a spatiotemporal trajectory index using spatial filling curves in the HBase distributed environment.(2)Adaptive load balancing based on multiple factors:Considering the extreme inclination of trajectory data,JUST+is proposed based on the JUST scheme to achieve adaptive load balancing of trajectory sets within the cluster,ensuring a roughly balanced distribution of hot and cold partitions within the cluster.(3)Compare the performance of JUST+with other solutions from various dimensions:From the perspective of read and write,comparing the read and write perfor-mance differences of three track index construction schemes,STR tree,JUST,and JUST+,proves that JUST+has better write performance and similar range query performance.From the perspective of load balancing,comparing the balancing effects of JUST and JUST+load balancing schemes proves that JUST+can adaptively adjust the cluster load and achieve read/write load balancing.For the trajectory range query scenario of bulk commodity logistics,this paper designs and implements a trajectory storage framework JUST+based on dimensionality reduction index and adaptive load balancing.Experiments are conducted using trajectory data sets from actual scenarios,and the results are compared with STR trees and JUST frameworks.The results show that the trajectory storage framework in this paper can greatly improve trajectory writing performance,outperforming the original JUST framework in range query performance,At the same time,it can better balance the load between node partitions between clusters.