Arquitetura híbrida de armazenamento no contexto da Internet das Coisas

• Autor(a) principal: Cavalcante Junior, Bráulio Lívio Dias
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFS
• Texto Completo: http://ri.ufs.br/jspui/handle/riufs/10665
• topic: Internet das Coisas; IoT; NoSQL; Armazenamento; Bancos de dados; Armazenamento em IoT; Storage; Databases; IoT storage; CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

IoT is present as one of the great areas of innovation in technology. Through it, it is possible to share information about the use of small devices, considered pervasive, that is, that are present in the daily lives of people without being perceived. Cloud computing has become a key component in developing IoT applications. As the number of devices increases, a large amount of data is generated. In this way, it is necessary to adequately treat the storage and access of this data in an optimized way. There are different ways of managing and storing data in IoT, including relational approach (SQL), non-relational storage (NoSQL), NewSQL, and large-scale models. In IoT, it is common to deal with unstructured data, most of which, besides applications, require high flexibility and frequent schema changes. For systems that require these types of requirements, previous studies indicate the use of non-relational storage, known as NoSQL. Databases of these types relax consistency and integrity constraints, provide horizontal scale, and optimize replication and access mechanisms. However, each data in IoT possibly requires a better type of specific NoSQL storage given a chosen criterion. Assuming that reading and writing performance is the main criterion for choosing the type of storage, it was necessary to identify which databases have the best performance of these operations for IoT data, which is one of the objectives of the study. However, prior to the evaluation, it was necessary to design a model to map the flow of data in IoT and the targeting of that data to the appropriate storage location. For this, the present work proposed a hybrid storage architecture for IoT. In order to validate it, an evaluation using the 3 main NoSQL databases were done to evaluate the insertion time and reading of scalar, multimedia and positional data, 3 of the main data types in IoT. After the evaluation, it was identified that the Redis database, of the key-value type, obtained the best performance for writing and consulting these types of data.