Previsão da capacidade de processamento em computadores pessoais utilizando ARIMA e assinatura comportamental

Abstract

Outlining the behavior of an important resource for a distributed environment can lead to a more assertive interaction between the manager and its suppliers and consumers, making its consumption more rational. The purpose of a Behavioral Signature is to trace the standard behavior of the use of one or a set of resources necessary to establish the capacity of a piece of equipment to serve the distributed environment to which it belongs. This work presents a proposal that allows the creation of a Behavioral Signature through the ARIMA model, based on the CPU consumption in personal computers, thus enabling a system that allows the prediction of this consumption through the continuous updating of this Signature. The developed methods were evaluated using the Akaike Information Criterion (AIC), in comparison with an automatic ARIMA model generator. The data obtained and their analysis showed superior results to the methods generated by the automatic model, presenting a higher level of assertiveness in this proposal of predictability and are treated in detail in this master's dissertation. Another motivator of this study is that the Behavioral Signatures obtained have a physical size that will enable the transfer to the manager module with low transmission overhead.