Analytical-stochastic availability model of multi-functional cloud infrastructure

Abstract

Purpose. Analyze stochastic methods for metamodeling of cloud infrastructures (СI) based on conceptual requirements for ensuring their readiness taking into account performance requirements, scalability of the computing resource, energy saving and elasticity of control of the components of the СI. The technique. The analysis of stochastic methods used to model the behavior and assess readiness indicators, the reliability of cloud infrastructures (СI) is performed. As a fundamental basis, the taxonomy of OO metamodeling is considered, which is based on the conceptual foundations of ensuring productivity, resource scalability, energy saving, and elasticity of management of the components of the cloud infrastructure. In addition to the well-known estimation methods using the apparatus of stochastic Petri nets, Markov chains are proposed to focus on the possibility of using semi-Markov modeling methods that contribute to improving the accuracy of evaluating the metrics of the quality of services provided (QoS). Results. The proposed taxonomy is a fundamental basis for modeling behavior and assessing the availability and reliability of cloud infrastructures. Scientific novelty. The problem of the lack of methodology for modeling the processes of СI functioning is solved, which is based on a single taxonomic basis of a comprehensive solution to the problem of assessing, analyzing and monitoring the level of readiness of cloud infrastructures. Practical value. The analysis opens up the possibility of using stochastic methods to ensure performance, resource scalability, energy saving and elasticity of management of the components of the cloud infrastructure. The considered methods of modeling can be applied to the selection of optimal architectural solutions in accordance with the established criterion of readiness of cloud infrastructures.