Uma abordagem baseada em métricas para explorar alternativas de esquemas de dados no processo de conversão de RDB para NoSQL

• Ano de defesa: 2020
• Autor(a) principal: Kuszera, Evandro Miguel
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Paraná; Dois Vizinhos; Brasil; Programa de Pós-Graduação em Informática; UFPR
• Palavras-chave em Português: Banco de dados; Banco de dados não relacionais; Banco de dados relacionais; Data bases; Relational databases; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO; INFORMÁTICA (40001016034P5)
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/32143

With the emergence of new applications, new requirements on storage systems have also emerged. Scenarios involving structured, semi-structured and unstructured data are increasingly common. Relational databases, widely used to store data from different applications, no longer adequately address all issues imposed by different scenarios. As an alternative, NoSQL databases have emerged, which are flexible in relation to the data model and designed to provide high scalability and availability. Relational databases and NoSQL databases will coexist for a long period of time and, as a consequence, new approaches to converting the relational model to NoSQL data models have been proposed. However, most of these approaches are aimed at converting relational data to a specific NoSQL data model and provide little support for customizing the conversion process, such as selection of fields, tables, instances, and other aspects related to the customization of the data schema produced. In addition, there are several ways to structure the data (or ways to define data schemas) when converting RDB to NoSQL. The choice of the appropriate data schema is not trivial and involves several aspects, such as the data access pattern, the desired level of data redundancy, the size of the resulting NoSQL database, the application maintenance effort, among others. This thesis defines an approach to convert and migrate relational data to document-oriented and column family NoSQL models, composed of an evaluation step of candidate NoSQL schemas. The approach uses directed acyclic graphs (DAG) to specify the structure of the entities that will be migrated to the NoSQL data model and also to represent the application’s access pattern (queries). To evaluate candidate schemas, a set of metrics and scores was defined, which aims to measure the coverage of the NoSQL schema in relation to the set of queries. As NoSQL schema and query are defined through DAGs, it is possible to perform evaluations and comparisons objectively. To evaluate the approach, we performed experiments involving RDB to NoSQL conversion scenarios composed by different candidate NoSQL schemas. The results of the experiments showed that the approach is effective to identify scenarios in which there is a greater effort to implement the queries, assisting the user in the process of selecting NoSQL schemas, before executing the data migration.