Availability Modeling and Analysis in a Virtualized Servers Network

Abstract

Virtualization has been considered as an inventive approach for modern information systems in the revolutions of cloud computing and big data. Virtualized systems have been taken into account in research in different granularities. In the trend of research on virtualized systems, this paper presents a new approach of virtualization in a data center. Our work aims to propose a network of virtualized servers incorporated with live VM migration to guarantee the localization of fault-tolerance. In order to maximize resource utilization and system availability, we introduce three rules of live VM migration inside the network. The proposed network system is modeled by Stochastic Reward Net (SRN) and analyzed by Stochastic Petri Net Package (SPNP). As an important measure of interest, system availability is our scope of analysis. The analysis results show new findings that are only revealed in a network view. Networking incorporated with live VM migration enhances remarkably system availability. Based on the analyses, we challenge the system developers to pay more attention on the core points of system.