Eficiência dos modelos de dados relacional, documento e grafo para o gerenciamento de dados geográficos móveis

Abstract

Increasing data volume, application diversity and mobility are the foremost characteristics of many current information systems. However, selecting the appropriate computational infrastructure is still a hard task for the designers of such systems. This work demonstrates how current performance evaluation techniques may not be ideal when comparing the performance of spatial database management systems. We consider mobile applications that involve constant spatial data traffic, such as querying for points of interest and map visualization. Our evaluation shows that a fair comparison requires specific workloads for each mobile feature -- which is not currently achievable by the industry's standard benchmark tools. We then compare technologies in relational (SQL), document-oriented and graph-based DBMSs (NoSQL). Overall, this study demonstrates that the one-size-fits-all benchmark methodologies are not optimal and may lead to a far from optimum system design.