Classificação de dados quentes e frios em bancos de dados em memória utilizando técnicas de Aprendizado de Máquina

Abstract

Abstract With technological advancement and the growing demand for processing massive volumes of data, the use the database has expanded significantly. Given this scenario, efficient systems based in main memory, such as In-Memory DataBase (IMDB), have emerged as prominent solutions. However, the increasing risk of data overflow highlights the importance of data management strategies. data, such as using Online Transaction Processing (OLTP), to move data between different storage systems. This study’s main objective is to develop an approach focused on Machine Learning (ML) techniques to identify and classify "hot" data and "cold" in memory-based databases. When prioritizing the efficient management of this data, we seek mitigate the data overflow problem. Furthermore, the study suggests possible strategies for using of identified data, such as storing "hot" data in RAM and allocating "cold" data on secondary subsystems, providing a basis for future studies to explore these data management approaches.