Arquitetura para análise e visualização de dados geoespaciais na Web

Abstract

Processing large volumes of geospatial data over the Web is typically implemented using one of two approaches: all processing on the server side or transfering data for processing on the client. In the first approach, is necessary to gradually increase the computational infrastructure, since each request initiates a process of high computational cost. The second alternative requires transferring large volumes of data across the network, with significant impact on the time to produce the result. This thesis proposes an architecture and presents a mixed implementation approach of an online service of geographical data processing. In this service, the server simplifies the stored data, with the intention of reducing the volume of information to be transferred to clients. Thanks to that, the clients work with data that is closer to what its operation needs and following that, completes necessary processing. The simplification process is applied only once, and the result can be reused in subsequent requests. The resulting service is more stable, more scalable, and performs better than previous approaches. This work includes the implementation of simplifiers and operations for map algebra where the input and output are matrices. Thus, the validation of the results obtained is done through a case study, developed in the software DynamicaEGO using the data from the Amazon Forest. In the study, the performance of the proposed hybrid approach is compared to the processing performed exclusively on the server. The result shows that is possible to perform, in an online platform, complex queries involving large volumes of data, with performance that allows multiple concurrent users.