Dimensionamento vertical automático de recursos em nuvens computacionais

Abstract

Cloud computing is a paradigm capable of providing processing, storage, network and other resources, for different types of applications. Cloud computing becomes popular by allowing the sharing of resource usage by different applications. This makes it attractive because clouds allow applications to operate on demand, reducing operating costs by application providers. Virtual machines provided by clouds encompass or host applications. With the use by these applications, virtual machines may require more processing power or memory resources. In case of saturation, situation where the virtual machine has bottlenecks in some of its resources, there may be loss of performance in the application. Therefore, there is a challenge of how to manage the resources of a cloud in order to meet the demands of different virtual machines and improving the allocation of these resources, in order to avoid the saturation of resource use. In this context, this work proposes a prototype of cloud scaler, which delivers on-demand resources to saturated virtual machines, or removes underutilized resources from virtual machines, without interrupting the virtual machine itself. To evaluate the proposed prototype, three case studies are made, the first with a synthetic load generator, the second with a web server under heavy load, and the third is the remote control of a robot in an intelligent space