Research On The Key Technologies Of Geographic Vector Data Engine Based On Multi-source Key-value Databases

With the rapid development of earth observing technology and the growing users’ demand for spatial data accessing, spatial database, on the basis of relational database, is faced with a series of challenges:highly concurrently read/written, data efficiently storaged/accessed and so on. Geographic vector data, as an important component of spatial data, has complex data structures, making it a greater challenge to be effectively organized and managed. Due to the difficulties faced by traditional relational database, NoSQL (Not Only SQL) technologies rise rapidly. Due to abandon of the ACID model, transactional consistency and paradigm constraints, NoSQL solves many problems faced by relational database. It has become an international leading edge and gained extensive attention and study.The paper is supported by the national science and technology infrastructure "Date Sharing Infrastructure of Earth System Science-Date Sharing Infrastructure of Yangtze River Delta". Under the guidance of the NoSQL, spatial database, formal languages and automata theory, it studies geographic vector data engine based on multi-source key-value databases. Key-value database is an important type of NoSQL and one kind of mainstream databases for big data management, with simple Application Programming Interfaces (APIs), fine expansibility. However, each different key-value database suit different applications with its unique advantage. In the application of key-value database into geographic vector data and its organization and management, it has been a problem to be solved of how to choose the appropriate key-value database in different scenarios, how to use unified methods to transparently access, and simultaneously maintain the advantage of underlying key-value database. This paper proposes a geographic vector data engine based on multi-source key-value databases. The paper is organized as follows:Firstly, under the guidance of the OGC (OpenGIS Consortium) simple features model and the storage mechanism of multi-source key-value databases, it designs a hierarchical organization model of vector Geodatabase, analyzes the storage method of metadata, indexes, features in different key-value databases, studies mapping mechanism between model of vector Geodatabase and physical model.Secondly, it designs highly extendable structure of key-value geographic vector data engine. The bottom of this structure is the accessing layer of multiple-source key-value database, according to characteristic of key-value database and data access mapping strategy, it shields the differences of heterogeneous key-value database and develops its unique advantage. The middle of the structure is accessing layer of vector data, combining the spatial index with the model of hierarchical organization, achieves querying and accessing geographic vector data by accessing layer of key-value database. The upper level of engine is the interpreter layer of GSQL (Geographical Structured Query Language). This paper studies GSQL interpreter based on formal languages and automata theory, following the process of lexical analysis, syntax analysis and semantic analysis.Finally, based on the above studies, the paper achieves an open, flexible and extendable geographic vector data engine prototype. This prototype system integrates several key-value databases, supports grid index, and provides spatial query language-GSQL automatic parsing. Compared it with traditional relational spatial database------Oracle Spatial, experimental results show that the multi-source key-value geographic vector data engine can take advantage of bottom key-value database, superior to traditional relational spatial databases in response speed and concurrency performance.