A Fault Localization Framework for Dynamically Provisioned Virtual Machines

In cloud computing systems, virtual machines that are provisioned in servers run for providing many services. Due to the flexibility of running virtual machines, the scale of cloud computing systems is becoming larger, which makes it difficult to detect the fault that causes failures of services. In this paper, we discuss a fault localization framework for dynamically provisioned virtual machines. Focusing on that Bayesian network has been used for the fault localization, we also use Bayesian network in our fault localization framework. However, the conventional methods address the static system configuration where servers and virtual machines are not changed. In cloud computing systems, the servers can be added easily and the virtual machines can be migrated to the other servers. In addition, requests to the services are changing, which sometimes needs scaling out or changing configuration of the systems. We discuss the research issues considering the above characteristics of the cloud computing systems.