A Hierarchical Model for Virtualized Data Center Availability Evaluation

Authors

Matheus D'Eça Torquato de Melo
Erico Guedes
Paulo Maciel
Marco Vieira

Abstract

Availability events like hangs, failures, and interruptions are usually unexpected and unpredictable. Due to the involving uncertainty, setting up a consolidated approach for availability evaluation is a hard endeavor. The availability evaluation becomes even more complicated in complex systems like cloud computing. Previous works fill such a gap using availability models. However, the majority of those papers proposes models for specific scenarios, which impairs their scalability and flexibility. In this paper, we present a scalable availability model for virtualized datacenters. From the model's design point-of-view, we provide: i) a hierarchical model based on Reliability Block Diagrams and Stochastic Reward Nets, and ii) an evaluation of the time to compute the model's results. We tested different architectures regarding system availability, capacity-oriented availability, and power consumption. As our model considers the fundamental premises of virtualized environments, it can be extended to specific scenarios. We also used a multiple-criteria decision-making process to highlight the best architectures for different users' needs, considering system downtime, capacity-oriented availability, and power consumption.

Keywords

Availability , Virtualized Data Center , Hierarchical models , Stochastic Petri Nets , Cloud Computing

Subject

Cloud computing availability modeling

Related Project

ATMOSPHERE: Adaptive, Trustworthy, Manageable, Orchestrated, Secure, Privacy-assuring, Hybrid Ecosystem for REsilient Cloud Computing.