Framework for testing the fault-tolerance of systems including OS and network aspects

This paper presents an extensible framework for testing the behavior of networked machines running the Linux operating system in the presence of faults. The framework allows to inject a variety of faults, such as faults in the computing core or peripheral devices of a machine or faults in the network connecting the machines. The system under test as well as the fault- and workload run on this system are configurable. The core of the framework is a User Mode Linux, which runs on top of a real world Linux machine as a single process and simulates a single machine. A second process paired with each virtual machine is used for fault injection. The framework will be supported by utility programs to automate testing and evaluate test results