Experiments and Analysis on Hypervisor-Based Fault Tolerance in Virtualized Cloud Environments

摘要

Virtualization is the key technology for the Infrastructure as a Service (IaaS) paradigm in Cloud computing. A IaaS provider operates a hardware system on the top of which the users deploy their computing infrastructures, implemented as Virtual Machines. Such Virtual Machines may be required to guarantee different service levels, in particular for what concerns scalability and Fault Tolerance. The paper focuses on a promising technique called Hypervisor-based Fault Tolerance. The analysis is based on a set of 'black-box' experiments on the proprietary VMware's technology called vSphere. The paper's contributions include a description of how the VMware Fault Tolerance system works and a description of how the performance of different types of applications (in particular I/O-intensive and CPU-intensive) are affected by the inclusion of such a Fault Tolerance mechanism.