Modelos para avaliação de disponibilidade orientada a capacidade de uma nuvem privada

Abstract

High availability is a significant requirement of applications hosted in clouds. It is possible to apply hardware and software redundancy to achieve higher levels of system availability. However, besides system availability, we also want to know system capacity to deal with the workload. We can use the Capacity Oriented Availability (COA) to obtain this information. The COA gives an estimate of resources available when the system is running. This paper presents a comprehensive analytical model to evaluate COA of private Clouds. We propose six scenarios with different architectures. All the architectures have the same three essential components: Front-End, PM, and VM. The presented results cover availability, COA and sensitivity analysis of model parameters. By using the model's results, it is possible to point out what components are more important for each studied metric.