For many years, network engineers had a dream. Their vision was that one day a network’s operations and management would be driven by its applications.

Today, with software-defined networking (SDN), that dream is finally becoming reality. ADVA Optical Networking recently played a key role in a major European multi-layer multi-vendor SDN field trial that was proof of this. The demonstration highlighted some of the key achievements of the past few months. As well as use cases like traffic optimization and multi-layer re-routing following cable breaks, the most significant breakthrough was the utilization of an SDN architecture in real-world production networks.

Juniper Networks and ADVA Optical Networking had already demonstrated packet optical multi-layer coordination at SDN OpenFlow World Congress in Düsseldorf, Germany in October 2015. In that showcase, Juniper’s router nodes and ADVA’s optical network were coordinated by Juniper’s SDN controller, NorthStar. The controller was aware of both packet and optical layer topologies. Based on this, the SDN controller was able to optimize routing decisions, such as traffic flows. The demo showed how several high priority packet paths could be routed on diverse optical links. In case of a cable break, the non-affected optical links would still be available and therefore overall network availability would increase. If the controller hadn’t been made aware of the optical topology this would not have been the case and all packet paths could have been routed via the same optical link.

So how is all of this possible? While the routers are directly connected to the controller, a kind of adaptation device is used for the “analog” optical layer. ADVA Optical Networking developed a so-called Network Hypervisor which delivers an abstract view of the optical network topology to the controller. The gives information on which links between router ports are available, associated shared risk link group information (i.e. “Are these links using the same cable?”) and latency data per link. As mentioned, the SDN controller is able to optimize routing decisions based on this. The interface between the Network Hypervisor and the SDN controller is an IETF standard conform REST interface based on a YANG model.

Optical latency information is another great example of how the network can be optimized to better support its applications. When I was in South America recently one of the network operators told me that sometimes their enterprise customers complain that network latency is suddenly higher after re-routing. I was able to explain what was causing it. The reason is that the packet layer isn’t aware of the optical topology and the latency values of the links. Therefore, it could happen that the packet re-routing process chooses very long optical links with extremely high latency values. SDN-controlled multi-layer coordination could provide the solution.

From an overall network architecture perspective, SDN involves applications communicating directly with business and service orchestration platforms. And these orchestration systems are connected to the SDN controllers. These architectures bring many benefits. The ones that get mentioned most often include the potential for increasing network agility, functionality, and availability while reducing production cost. However one of the most important drivers for network operators is multi-layer coordination in open multi-vendor environments.

Thinking about the impact that these breakthrough technologies are finally starting to have on the industry puts my mind back to 2004 when engineers in Germany conducted a field trial related to this. It was called Vertically Integrated Optical Testbed for Large Applications (VIOLA). One of the project’s goals was the “development and test of software tools for the user-driven dynamical provision of bandwidth”. Now, does that sound familiar?

It’s amazing to think how far the story has progressed. And the next chapter is already beginning. As we speak we’re preparing for the next trial, not in research labs but in an actual deployed network, this time in the US.

So it seems that, while Düsseldorf was another small step, the deployment of SDN-controlled multi-layer coordination in production networks will be the giant leap.