NFV‐FD: Implementation of a failure detector using network virtualization technology

Summary A failure detector (FD) is a classic distributed systems service that can be used to monitor processes of any network application. Failure detectors provide process state information: A process is reported to be either correct or suspected to have crashed. In this work, we propose the implementation of a process monitoring service using Network Function Virtualization (NFV) in an OpenFlow network. The NFV is a technology that uses software virtualization techniques to drastically reduce the cost of deploying and managing network functions that are usually available as middleboxes. We describe NFV-FD, a network service that provides information about the state of both processes and links. The NFV-FD relies on an OpenFlow controller from which monitoring information is obtained. We investigate different ways to implement NFV-FD. We evaluate the performance impact of monitoring by implementing the virtual function both within and on a separate host from the controller. Then we compare an implementation based on a traditional virtual machine with another implementation based on containers. Experimental results are reported not only in terms of performance, ie, the amount of resources required by NFV-FD, but also for the quality of service provided by the failure detection and notification functions as they are used to deploy a reliable broadcast service.