Modelagem de desempenho, dependabilidade e custo para o planejamento de infraestruturas de nuvens privadas

This work presents an integrated solution composed of a methodology, methods, representation models, optimization models and a tool, for the planning of private cloud infrastructures, considering aspects of performance, dependability, performability and cost. The proposed optimization models are based on the GRASP metaheuristic for generating private clouds infrastructure scenarios. The model for performance and cost scenario generation allows the creation of cloud infrastructure scenarios with different software and hardware configurations and the model for availability and cost scenario generation permits the creation of cloud infrastructure scenarios with various redundancy mechanisms assigned to the components of these infrastructures. The proposed representation models are based on stochastic Petri nets, reliability block diagrams and mathematical expressions for performance, dependability, performability and cost evaluation of cloud infrastructures. Performance model allows the evaluation of the impact of varying the workload submitted to cloud computing with different software and hardware configurations. This performance model permits the calculation of the response time and resource utilization of the cloud infrastructures. Availability models allow the evaluation of the effect of assigning different redundancy mechanisms to the cloud components. The availability models permits the calculation of the availability and downtime of the cloud infrastructures. Cost models allow the evaluation of the acquisition of equipment of the IT system, acquisition of redundant equipment and software, equipment replacement, acquisition of software licenses, technical team and maintenance team costs. Performance, dependability and cost metrics can be combined to select cloud infrastructures that meet the client requirements. The proposed tool provides the automation of the integrated solution. Three case studies are presented to implement the proposed solution. The first case study presents the evaluation of the impact of the workload variation assigned to the Eucalyptus platform with different software and hardware configurations. The second case study presents the evaluation of the effect of the assigning redundancy mechanisms to the components of the Eucalyptus platform. The third case study presents the evaluation of the impact of the occurrence of fault and repair activities in the performance of the Eucalyptus platform.