Query Processing Of In-Memory Spatial Data Base Based On Cluster Environment

The development of spatial data acquiring technologies continuously increases spatial-temporal resolution of spatial data,which leads to rapid expansion of spatial data size.Meanwhile,location-based services explosively emerge,for which the massive spatial data can well support.However,high concurrent I/O operations on huge volume of data bring great challenge to spatial database.Traditional spatial databases are incapable in this scenario,due to restrictions brought by hard disk I/O speed and ACID.High performance clusters and RAM of large volume provide new possibilities at hardware level.This thesis investigates how to effectively store and query vector data within distributed memory based on NoSQL database and high performance clusters.In-memory spatial data model,high performance concurrent processing,and spatial extensions to NoSQL database are discussed.Appropriate scenarios for proposed methods and software are illustrated.The experiment results provide solid basis for subsequent implementation of spatial database based on in-memory database.A hierarchical vector data model optimized for distributed memory is designed and implemented.A spatial data partitioning strategy based on Geohash is proposed,which ensures that geographically near features also lie on nearby storage.An in-memory spatial database named Spedis is implemented by extending Redis with spatial functionalities,of which querying performance is compared to PostGIS.The experiments suggest that Spedis has better performance than traditional spatial database,especially in high concurrent scenarios.Spedis is integrated into the high performance geographic computing platform,HiGIS,to verify its capability to speed up management,query and processing of spatial data in high-performance geographic information system.