In order to realize the vision for software-defined networking (SDN), applications must be liberated from the underlying network details to attain the degree of openness operators are seeking for cloud, data center, enterprise, or carrier networks. This will only be possible through abstraction – enabling applications to yield the portability and agility the industry has been striving for.

Ultimately, abstraction allows controller developers to publish a common API framework for network and business application developers to utilize. Application developers are then spared from the complexities of integrating their applications with each controller, and operators would be afforded the choice they are seeking to avoid vendor lock-in, while the need for a proliferation of “glue” software that has limited incremental value is also removed.

Of course, the real world is never that simple, not when it comes to networking. Nowhere is the bar higher than for application-to-network integration. However, industry groups, led by the Open Networking Foundation (ONF), have leapt at the challenge, forming the Northbound Interface working group. The open-source community was motivated as well, with both OpenDaylight and ONOS, open-source controller frameworks devoting significant resources to the controller-to-application interface.

Individual companies also responded with innovative work, such as Plexxi’s Affinity API framework (contributed to OpenDaylight). Early SDN innovators have developed their own northbound API to ship early SDN products, and their progress has helped simplify application and network management integration across their own disparate platforms and bootstrap important work in the ONF, OpenStack, OpenDaylight, IETF, etc.

However, the common API framework serving the needs of operators, applications developers, and controller developers remains elusive.

A recent gathering hosted by HP Labs in Palo Alto offered encouraging signs for the broad industry collaboration required to address the industry’s most significant challenges. About 70 individuals participated in the Intent-based Summit, the name referring to a an application’s desire to express its intent, independent of whatever network infrastructure on which it will be running. David Lenrow (HP), chair of the ONF Northbound Working Group, contributed a companion article in SDxCentral that describes intent-based networking in more detail.

The Summit was among the first meetings of its kind, attracting broad representation across SDN and NFV industry groups, standards, and open source projects. The primary goal was to initiate a rallying cry for the industry to set aside their different approaches, competitive positions, and internal objectives for the benefit of the network operators that they are serving.

No one expected an issue this complex and important to be addressed during the day-and-a-half Summit. However, the sense of collaboration, urgency, and openness did surprise some, and after the major players shared their story, many misunderstandings were overcome.

Among the outcomes:

• Two distinct camps emerged, which vary in the level of abstraction. Application-centric organizations consider application-level interactions, such as how VMs communicate with what policy. Network-centric intent addresses relatively lower-level connectivity.
• Northbound APIs are all relative, and there will be multiple layers to address
• There was general agreement on the problem statement, referred to as the multiple-writers problem. In an SDN architecture, multiple applications can alter the network forwarding behavior in ways that sometimes conflict. In many controllers, there is not only conflict resolution, but only rudimentary conflict detection at best.
• A general concern expressed was the lack of a unified model that governs the semantics and operations applications they can expect from the various controllers
• More needs to be done to develop a general approach, and a unifying forum is required to make progress, Finally, OpenDaylight appears to be emerging as an important forum to encourage a collaborative dialogue

The broad range of use cases being considered for OpenDaylight means additional work is recommended to further define the target use cases and subsequent requirements to guide the initial work. Six OpenDaylight members banded together to propose a new project to address the work: the Network Intent Composition (NIC). The proposal passed, and the work is already progressing.

The ONF is attempting to coordinate the activities actively involving the other constituents as well, including ONOS, OpenStack, OPNFV, and of course all OpenDaylight contributors. Contributors are ramping up one of the most ambitious efforts to date to unify the community and realize the level of collaboration necessary to tackle the most challenging problems -- is there a more pervasive problem than the Northbound?

In the near term, it is unlikely that there will be a single intent-based networking approach that can address the diverse needs across the SDN community, nor does anyone expect a single open, Intent-based API that is sufficient for all applications and SDN controllers. If we do not converge on a single approach, the value may be limited.

But the intent for intent is clearly on the right track and a departure from the traditional approach, where groups struggle and sometimes battle to refine their charter, incurring delays and ill will. Achieving application portability in SDN remains a challenge, especially in an open and multi-vendor environment. Intent-based networking offers a sound starting point but will no doubt evolve, guided by a positive intent from many with no looking back.