Enabling Fast Failure Recovery in OpenFlow networks using RouteFlow

Abstract

OpenFlow provides a protocol to control a network from an external server called controller. Moreover, RouteFlow presents a framework to run Internet routing protocols in OpenFlow networks by running them in virtual machines or containers. The problem is that OpenFlow networks running RouteFlow do not recover fast from a port failure (e.g., port down event). The failure recovery time is dependent on user configurable parameters and is in seconds. To overcome this problem, we implement a solution in which a port failure of a physical OpenFlow node is detected immediately in its corresponding virtual machine and an immediate action is taken by the routing protocol. Therefore, once a routing protocol running on the corresponding virtual machine detects this failure, it broadcasts the failure in the network and a new failure free path is immediately configured over the OpenFlow network. We implement the proposed solution in an OpenFlow controller and test it over single autonomous and multiple autonomous system scenarios (including OpenFlow and non-Openflow scenarios) of the Internet emulated on the virtual wall testbed of the Fed4Fire facility in Europe. The results show that an OpenFlow network can recover from a failure in a short time interval using the proposed solution.