Predição da disponibilidade de recursos computacionais a partir do uso de médias móveis em séries temporais

Abstract

This work has as general objective the analysis of time series generated through collections of the percentage of CPU usage in microcomputers aiming at the identification of the equipment behavior. For that, it was necessary to understand the characteristics of these time series, and from them, provide the study and the application of methods involving moving averages in the collected data in order to obtain a behavioral signature of the monitored equipment. This behavioral signature must, throughout successive analyzes, be compared with the actual behavior of the equipment and with this, identify the margin of error involved in the process. The application of these mathematical methods will enable the modeling of a computational resource management system (CRM) based on the predictability of their occupancy rates, allowing and adapting their use with the highest possible success rate.