Analisando configurações de provisionamento automático na nuvem: uma abordagem baseada em simulação.

Abstract

The fluctuating demand of an application makes the automatic provisioning of resources a complex activity, as it is necessary to ensure that the infrastructure supports the workload without impairing the application’s performance, while avoiding unnecessary costs. For this, an automatic provisioning policy must be carefully configured, facing the complexity of the large number of parameters involved. Moreover, effective strategies for one application may not be suitable for another, and inappropriate choices can cause delays in resource allocation during peak demand, negatively affecting the user experience and increasing operational costs. Such choices can be even more challenging due to the inexperience or lack of knowledge of the professional responsible. Thus, this study addresses the complexity in choosing appropriate settings for automatic provisioning, considering that studies show that this choice is made through intuition or limited tests in the production configuration. Our proposal is a methodology accompanied by tools to guide SREs in the efficient choice and configuration of automatic provisioning policies, which allows the user to simulate different policies and/or configurations and analyze the behavior of resource offering through metrics from the state of the art and others developed in this study. The methodology along with the analyses allow observing the behavior of demand and resource offering through modifications in the configuration, in addition to facilitating the replication of these analyses, whether for academic or industrial purposes. This approach allows for testing before deployment in a production environment, providing insights for more precise decision-making.