Scalable Management of Spatial Data

点击翻译按钮获取中文摘要

This project falls within the EPSRC Information and Communication Technologies (ICT) theme and theInformation Systems research area.In recent decades, the volume of spatial data (e.g. location, routing, navigation) is increasing dramaticallyacross the world. In this project, we are motivated by the fact that conventional data management andprocessing technologies are becoming inadequate for analyzing large-scale spatial data and there is anemergence of novel approaches. Our overall objective is to investigate the above-mentioned problem andgive solutions based on a crossover of Distributed Processing and Factorized Databases.Aims and ObjectivesInvestigate the theoretical foundations of factorized spatial data management.Spatial databases are optimized systems for storing and querying data that represent objects denedin a geometric space. As a result, they require additional functionality to be intergrated into theirquery languages in order to process spatial data types efficiently.We are going to accommodate recent development on factorized databases to evaluate typical queriesfor spatial databases. Factorizing spatial queries will enable us to avoid redundancy and as a resultachieve better performance and possibly asymptotical lower complexity.Optimized distributed processing algorithms for spatial data analysis.Our methodology will combine the developments on the algorithmic design of parallel and dis-tributed methods with the principles of factorized databases to lower the communication cost insuch distributed algorithms. We are going to use the Massively Parallel Computation (MPC) model,to analyze our results.Develop distributed algorithms for clustering, classication and pattern recognition of spatial data.Clustering, classication and pattern recognition tasks are very common in spatial data analysisbecause they provide a way of generalizing information retrieved from a database.We will propose new state-of-the-art algorithms to compute these tasks on factorized representationsof spatial data and compute them in parallel and distributed way.Distill practical lessons from applying the novel algorithms on real-world spatial datasets.We rely on our industrial partner Ordnance Survey to provide us with real-world spatial datasetsand liaise with them to nd natural ways to disseminate the outcome of our research to theirdevelopment team.Novelty of Research MethodologyFactorized databases are a fresh look at the problem of computing and representing results of relationalqueries. Consequently, our research methology is novel, since we are going to explore the special charac-teristics of spatial relational queries under the scope of the factorized representation and computation ofqueries over large spatial datasets. This is a novel research topic that has not been investigated so far.Alignment to EPSRC's StrategiesThis project aligns to EPSRC's priorities for data enabled decision making. This is because it aims tomake contribution on new methods to support people make decisions in a world that is becoming evermore data rich.CollaboratorsThe datasets will be provided by Ordnance Survey, the premier UK organization that produces topo-graphic maps. They are interested in this research and we expect collaboration with them on under-standing challenges and existing solutions for managing spatial data.