In order for service providers to successfully make the transition to the New IP, they must first deal with the huge challenge of creating fully programmable networks that successfully integrate a diverse mix of networking devices under one seamless management system.
For operators building out their New IP networks in greenfield environments -- similar to data center networks -- they can leverage software-defined networking (SDN) controllers and infrastructure devices that have already been optimized for the New IP ecosystem. However, most operators will have to realize their goal of creating a fully programmable network while still utilizing their existing infrastructure, which requires multiple methods of OSS integration.
In most service provider networks, networking devices can be segmented into one of three broad categories:
Three categories of networking devices in service provider networks.
Traditional networking devices: Most existing network infrastructure devices (e.g., DSLAMs, optical transport equipment, Ethernet switches, routers, etc.) utilize CLI or SNMP for programming. While these devices can certainly be programmed, the challenge is that every device requires a unique integration and there is not always a common provisioning model from different vendors, let alone the same vendor. This results in time-consuming integration efforts that delay time-to-market for new services.
In recent years, most operators have made the transition to allow vendor network management solutions (NMS) to serve as an abstraction layer between their orchestration systems and these infrastructure devices. Specifically, if the vendor NMS provides modern web services APIs that are service-aware, then the NMS can handle the heavy lifting of translating a simple, device-agnostic service request command from the higher layer OSS into the appropriate combination of CLI and SNMP commands.
The benefit of this approach is that OSS integration can be implemented with a fraction of the effort of direct CLI or SNMP programming from the service provider OSS -- where every device or service type requires unique integration effort -- while still offering the benefits of network programmability. The challenge is that service providers are more reliant on the vendor NMS solution which may not offer the real-time programmability of more modern SDN-based implementations. Depending on the operator goals and device type, these may not be major issues.
Structured, programmable APIs: Ideally, all networking devices will have well-defined, open APIs that are generated off a data model to create a common device provisioning model across multiple vendors. In the residential CPE space, TR-069 managed devices and DOCSIS cable modems have offered these well-defined data models, easing multi-vendor integration.
The same cannot be said for many traditional networking devices. Many operators would prefer to jump straight to having devices programmed via OpenFlow from an SDN controller, but for broadband access and optical transport equipment, these commands are not defined for these product groups. However, modern infrastructure devices are starting to integrate service-aware, open APIs directly into the network elements. This greatly simplifies integration of these devices into vendor-provided or vendor-neutral NMS solutions. Modern protocols for provisioning these types of devices include Netconf and REST.
SDN and NFV environments: These devices are optimized for SDN controller environments, preferably using OpenFlow for flow programming and a protocol similar to NETCONF for device configuration. As a result, adapters can be easily generated for vendor-neutral SDN controller environments. These devices also maximize the use of NFV to create more software-centric networks with minimal proprietary hardware. Currently, packet switching devices are the most ready for this environment, but efforts are underway to further standardize this approach across a broader range of devices.
As with many other network decisions, there is typically no one-size-fits-all approach to solving the challenges of creating a programmable network to meet the demands of the New IP. However, with the appropriate planning, operators can still realize their service agility goals with their existing infrastructure while reducing OSS integration costs, reducing integration timelines and integrating all devices into a common orchestration layer. Operators should also pay much more attention to the level of programmability offered by their networking devices and how progressive they are from a programmability perspective.
— Robert Conger, Assistant Vice President of Carrier Strategy, Carrier Networks Division, ADTRAN, special to The New IP