Um Recomendador de Alocação de Recursos em Computação em Nuvem usando Algoritmos Genéticos e SVR.

Abstract

The allocation of resources in Cloud Computing has been done in a reactive way, what can generate service guarantee failures and the charging of idle resources. In order to alleviate these problems, the objective of this work is to present a predictive resource allocation solution, in the form of a Configuration Recommender, using Support Vector Regression (SVR) and Genetic Algorithms (GA). As a case study, machine learning applications based on the Weka tool are chosen. This arrangement is used to compute application execution time and recommend a viable and valid configuration of cloud resources, based on the estimation of execution time and costs. The results show that the predicted times had an acuracy of 94.71% in relation to the real ones, thus leading to an efficient estimation of time and cost. In some cases of execution in the cloud environment, there was a reduction of time and cost of 38.8% and 45.62%, respectively.