Instantaneous recovery of unicast connections in transport networks: Routing versus coding

The importance of reducing recovery time and latency, e.g., by adding redundancy to the connection is increasing in today's communication networks. Thus, in this paper we investigate dedicated protection approaches which completely avoid control plane signaling and switching matrix reconfiguration when a network failure occurs, i.e., provide quasi instantaneous recovery from failures. As these approaches may require a huge amount of redundancy, we introduce a dynamic routing framework, called General Dedicated Protection (GDP) which provides optimal resilient capacity allocation against multiple link failures, both for routing and network coding. The computational complexity and a theoretical lower bound for GDP will be presented, too. We show the efficiency of the GDP framework through thorough simulations. Finally, we demonstrate that the necessary modules for GDP can be easily adopted in Software Defined Networks (SDNs) with the help of the OpenFlow protocol. In our proof-of-concept implementation we are following the Network Function Virtualization (NFV) approach, which allows the practical deployment of GDP in transport networks.