Uso de aprendizado de máquina e métodos de análise de dados para predição de desempenho em bancos de dados

Abstract

Technological advances in recent decades have generated an exponential growth in the volume of data in contemporary society. Coming from different sources, data drives the use of different storage systems. A well-known problem, which is frequently reported by the Information Technology area, is the constant decline in database performance. This is considered a big problem that could be fixed and make data processing processes more efficient. In this sense, having mechanisms that detect possible drops in the performance of a database would be a way of reducing the analysis and identification time of these drops in the database's performance by the analyst. This work aims to approach the use of machine learning algorithms to predict performance problems in relational database servers, using the database manager IBM DB2 on operational system Linux. Database performance was collected and tabulated, and a case study was conducted, evaluating different machine learning algorithms. The results were promising and indicated that it is possible to identify performance drops in the operation of a database server.