Hybrid network architectures offer both circuit-switching features (like traffic isolation, guaranteed bandwidth, and limited latency) and packet-switching features (like high flexibility and better resource utilization), enabling network operators to deliver highly flexible connectivity services with guaranteed service levels at competitive prices to their customers. A variety of different combined optical packet/circuit solutions already exist and are often used in hybrid network architectures.

The authors of this paper take a software-defined networking approach and propose an extension to the OpenFlow protocol to properly support hybrid networks. The network control plane is totally decoupled from the data plane, and most of the intelligence required to manage the network is concentrated in an extended OpenFlow controller. Moreover, the proposal includes a modular design of the node forwarding functions, separating local tasks from hardware and technology-dependent operations. In their proposal, an OpenFlow agent implements the standard northbound interface to the controller, while a logical forwarding element (LFE) performs any abstract forwarding tasks. For each type of switching and transmission technology implemented in the data plane, a physical forwarding element (PFE) is in charge of hardware-specific forwarding operations. For hybrid networks, typically two different PFEs are needed: one for packet forwarding and one for circuit switching. Finally, the authors describe the extended flow table matching rule format and the extended set of flow actions format.

The proposal was tested in a virtualized testbed, and the authors show that performance starts to deviate from ideal behavior at 30,000 packets per second (pps). The OpenFlow classifier and the optical switch emulator were identified as bottlenecks.

The main contribution of this paper is that it shows how software-defined networking approaches can be applied to hybrid network architectures. The target audience includes researchers and networking practitioners interested in new concepts and results within the area of software-defined networking.