Building the case for Software Defined Networking for metro networks

Michael Kennedy, ACG Research


Software defined networking (SDN) architecture that has its origins in data center networks is a promising approach for service providers' metro networks. It offers a means of reducing costs (capex and opex) and as a vehicle for increasing service velocity.

The Open Networking Foundation, a nonprofit consortium dedicated to establishing a standardized version of software defined networking, characterizes SDN as a network architecture where network control is decoupled from packet forwarding and is directly programmable. Other characteristics of SDN include logical centralization of network intelligence and state, and abstraction of the applications from the underlying network infrastructure. OpenFlow, a standard communications interface defined between the control and forwarding layers of an SDN architecture, is an important part of the Open Networking Foundation's SDN initiative. The motivation behind OpenFlow is to free service providers from dependence on their incumbent vendors and enable them to have a greater role in architecting their networks and service offerings. The SDN and OpenFlow initiatives are motivated by the successes that data center initiatives like compute and storage virtualization have achieved. In recognition of this, SDN approaches also are referred to as network virtualization.

The legacy TDM/PDH operational model embodies many concepts similar to SDN architecture. Network control is centralized while the transport equipment provides static routes and traffic flows across the network. A simplified operational model that limits the complexity of each network element is employed. This reduces the skill requirements of the field service technicians responsible for maintaining and operating the network. This strategy also greatly reduces the cost of the transport equipment because it has no costly or complex route control components.

Metro network transport networks are making a slow transition from TDM/PDH technology to packet technology. However, they must necessarily maintain the same operational model in order to control costs while accommodating high traffic growth rates. Affordably meeting wireless operators' mobile backhaul requirements is one well publicized technology issue.

There are opportunities to apply data center successes to the metro transport network. One SDN concept is to abstract the applications from the underlying infrastructure. This supports improved automation and management and separates the orchestration and provisioning of systems and applications from the network details. (Orchestration is the stitching together of software and hardware components to deliver a defined service.) In a recent study of cloud technology, I found that this reduced service delivery times from months to 15 minutes.

Virtualization also has been found to increase the flexibility and speed of deploying resources to meet changing requirements. This also reduces costs. I found a 35 percent TCO reduction due to data center virtualization in a recent study.

SDN approaches, though not necessarily OpenFlow itself, are being deployed as a means of reducing the cost of the metro network and improving service velocity. Cisco's (Nasdaq: CSCO) nV technology--though certainly not open--is one early effort to create a virtual metro network. It reduces costs by 70 percent when compared to today's router-based networks. The cost savings is achieved through much simpler and lower cost devices at the network's access and pre-aggregation layers. Ericsson (Nasdaq: ERIC) also has announced a network virtualization strategy with similar cost savings when compared to a router-based network.

A number of other metro network virtualization efforts are underway. Most of the ones I am aware of use MPLS-TP, a standardization project with ITU-T and IETF involvement, rather than OpenFlow as the interface between a centralized control layer and the forwarding layer.  MPLS-TP relies on static, pre-programmed routes rather than dynamic ones, just as SDN envisions. The control layer can be manual or software based as in the SDN architecture. The economic benefits of the MPLS-TP based network include simpler and lower cost access and pre-aggregation network elements and associated lower operations expense.

OpenFlow-based solutions for the metro network are most likely to be championed by those vendors that specialize in data center switching, while vendors that focus on service provider networking accounts are likely to favor MPLS-TP for virtualized network solutions.

Service providers appear to have a strong business case for deploying virtual networking (SDN) solutions in their metro networks whether based upon OpenFlow, MPLS-TP, or some other control to forwarding layer interface. However, it does not seem likely that the Open Networking Foundation initiative will actually open up the networking infrastructure market to a lot of new competition. The need to back up product sales with large-scale systems integration services and supply chain processes favors the large incumbent vendors. However, these initiatives may reduce the average selling price of individual systems by more than 50 percent. Therefore, unit sales must increase substantially to maintain current revenue levels. Strong demand for smart phones, tablets, M2M devices, wireless apps and cloud solutions may well provide the demand needed to offset the likely decline in average selling prices.

Michael Kennedy is a regular FierceTelecom columnist and is Principal Analyst at ACG Research. He can be reached at [email protected].