Avaliação de modelos de agrupamento para detecção de comportamento de aplicações em termos de demanda e uso de recursos.

Abstract

The use of application observation techniques has become increasingly common. Observing an application generates important data about its functioning and the infrastructure where it is inserted. Analyzing the behavior of applications is a key element that allows understanding and provisioning computational resources, optimizing the use of the infrastructure on which the applications run. Although there is behavioral recognition of applications in relation to the use of computational resources based on human decisions, the detection of high and low memory consumption behaviors, for example, through predictive models is still not very common, which opens opportunities for studies in this area. The present work proposes to detect application behaviors from different clustering algorithms. The results show that it is possible to detect each behavior to facilitate understanding and efficient allocation of computing resources.