Arguably, software-defined networking, or SDN, needs three things to succeed: First, it has to maximize its reward-to-risk relationship to promote adoption; second, it must present an evolutionary track that facilitates early adoption and allows operators to build on their early commitment; and, third, it has to make a name for itself that stands for something simple and cohesive.
Luckily, a solution can be found in service operations, by which I mean service lifecycle management as the set of software processes that relate to each stage of a service -- from its original customer order, through its deployment, to running live and finally to tearing it down. As defined, service operations can help SDN meet all three prerequisites for success. More to the point, SDN success could be difficult without it.
Let's start at the top to acknowledge one of SDN's big challenges: an imprecise definition. SDN has three distinct models:
- The first SDN model is the classic OpenFlow SDN developed through the Open Networking Foundation.
- The second is the cloud-driven overlay model popularized by Nicira, which is now VMware NSX. This model is represented by the exploding software-defined WAN (SD-WAN) market.
- The third SDN model is the API-driven approach best known in the form of Cisco's Application Centric Infrastructure (ACI). Cisco ACI adds application control to current network technology, which gets around the need for a forklift upgrade.
By dividing up the SDN space, these models have complicated the goal of SDN adoption. All of these models rely on one thing, however: service operations efficiency.
Service operations benefits to promote SDN
The three existing SDN models have different, low-level management approaches, but none by themselves would affect overall service lifecycle management to improve operations efficiency or service agility -- the two benefits that operators accept as credible SDN drivers. The lack of overall service lifecycle management improvement has pushed network functions virtualization (NFV) to embrace an orchestration model that has a broader effect on automating service lifecycle management tasks. A broad, high-level service operations model similar to the NFV approach could help SDN in two ways.
- The first improvement would guarantee SDN's benefits. Service operations or service lifecycle management can include all network and service operations tasks. It is based on an orchestration-driven automation strategy with a scope that can cover all possible operations costs and remove barriers to an agile response to competition or opportunity. If SDN is wrapped in an automated service lifecycle management framework, it can offer the best possible business case, which includes capital savings, operational savings and new service revenue potential.
- The second improvement is NFV's vision of service lifecycle automation that supports both legacy and SDN connectivity. The emerging approach to service lifecycle automation represents both virtual and real devices and networks as functionally defined abstract objects. This abstraction means it is possible to adopt SDN where aging infrastructure, fast-moving competitive pressure or new revenue opportunities dictate -- without dividing the network into new and old or complicating operations. These abstractions can represent any of the three existing SDN models, making it less critical to converge on a single approach.
Service operations automation affects each SDN model
Orchestration-based service lifecycle automation can address everything SDN needs to succeed. But the scope of the approach and its ability to embrace current network infrastructure also delivers many of the benefits without requiring an SDN evolution. If you can automate legacy device services, why not just do that and avoid the cost and risk of a transition?
Service operations may be the key to having SDN break out of its function-based abstraction to shine on its own. We know from the experience of giants like Cisco that SDN can radically improve the efficiency of both metro and transport networks. But to gain the most from SDN, you have to adapt its unique route-control capabilities to work with traditional IP and Ethernet services. This opens a new dimension in service operations -- one that could eliminate the risk of increased operations costs using a SDN-centric networking model or the risk of service problems.
Google's SDN approach lets an operator build what looks like a router core, but is in fact a virtual SDN core. A boundary function does the adapting between SDN and IP. Inside that boundary, SDN is used to improve trunk utilization and handles problems by providing controllable failover. This same model could be used by network operators and even by enterprises if service management was made efficient. The SDN-in-a-wrapper approach fits the service operations vision now evolving, and it would make the transition to automated service operations easier.
The overlay SDN model could also benefit from model-based service operations. With overlay SDN, including SD-WAN, there are two network service layers: the SDN overlay itself and the physical network, which is built using IP tunnels with MPLS, Generic Routing Encapsulation and Ethernet. Model-based service operations can represent both service layers and define how quality-of-service and service-level agreement parameters from the top are related to traffic management below. This automated coupling of service management from both service layers reduces operations effort and error, which is critical to control costs.
Even the simplest SDN model, embodied in Cisco's ACI, can be made more efficient through service operations automation. Effective implementation of this kind of SDN is almost always based on policies that must be defined and distributed throughout network infrastructure. Policy distribution isn't new, but it's rarely coupled directly to network or service management. Service operations through model-driven orchestration like that of NFV can provide the link between policies and network behavior, as well as between application services like quality of experience and policies. This link can significantly improve the way service users are supported, even to the point of facilitating automated customer support portals.
The concept of function abstraction inherent in service operations modeling is the only way to let SDN embrace the current state of the network, while at the same time breaking free from current network and service model limitations. The future of SDN lies in improving services, not just reworking the services we already have. A flexible service operations approach enables SDN to meet that future head-on to grow to its full potential.
For SDN success, virtualization and orchestration are key
Apply software automation to service lifecycle management
Service automation transforms more than SDN