Desenvolvimento de ferramenta para migração bidirecional e avaliação comparativa de desempenho entre bancos de dados relacionais e NoSQL

Abstract

Technological advancements have generated a massive amount of data in company databases, which can result in slow and costly processing for data reading and analysis. In this context, this project aimed to develop a data migration tool from MySQL to MongoDB and vice versa, with the goal of migrating a database from a relational database to a non-relational database. The tool successfully met all the expected expectations and was compared with the WithData tool to analyze if it migrated the data correctly as the WithData tool does, thus confirming that the tool correctly migrated the database in both ways. The study uses a MySQL database containing 80,000 records and performs the migration to the NoSQL MongoDB database. The research compares the functioning of the two databases, examining data storage, time required for insertion, updating, and search methods. Since the YCSB tool provides tests with the quantity that the user needs, the tests started with 1000 data 1000 times, followed by 25000, 100000, up to 1000000, generated from the database containing 80,000 records as mentioned earlier, in order to evaluate if the non-relational database demonstrates greater response agility. The main objective was to provide an analysis of which database model is the best option, considering performance in terms of processing time. The test results were initially similar, with almost identical times. However, as the tests progressed, it was possible to notice that MongoDB started to stand out, presenting better results than MySQL. MySQL ended up taking much more time to process the tests and processed significantly fewer operations per second. While MongoDB processed nearly twice the data of MySQL, it took much less time to process the same amount of data. Thus, it can be analyzed that the tests obtained clear results, with MongoDB standing out in relation to MySQL, presenting better times and processing many more operations per second than the relational model. There were only a few query tests where the performance of both was practically similar, but overall, MongoDB led, showing superior performance.