SDN will help bring an end to silos, cut delays and improve network utilization.
Recently, my colleague Paul Littlewood and I contributed an article to Global Telecoms Business discussing how SDN can eliminate network silos and can help improve network utilization. The complete article can be accessed here on page 64 with a short excerpt below. 
Imagine the savings that could be achieved if networks were operated with a unified model spanning all layers. This toolkit would have detailed knowledge of the whole network – end to end and up and down technology stacks. Network utilization could be optimized and network issues automatically corrected with minimum impact on services. The dynamic abilities of modern network elements at all layers would be effectively employed. To an operator the network would appear as a cohesive dynamic entity; gone would be the operations silos of today that result in delays, errors, and inefficiencies. This dream is now becoming a reality, thanks to Software-Defined Networking (SDN).
With new technologies being specified by the Open Networking Foundation (ONF) and in development by leading equipment suppliers, this vision is within reach. The starting points in many networks have a lot of the ingredients. Currently, large networks typically comprise a number of autonomous layers employing varying protocol suites and forwarding mechanisms. Separating the layers facilitated the myriad services on typical networks (e.g., the Internet) with widely varying requirements, and ensures capital invested in networks is efficiently used. But the silos often get in the way of efficiency; even networks comprising only a couple of layers can benefit from integration.
The drive towards convergence has been happening in networks for a number of years. Witness the widespread deployment of packet-optical transport systems, or the integration of DWDM interfaces into Ethernet switches and routers. However, one area where convergence has not happened is in network control.
Each layer has its own independent control plane, which generally is not coordinated with control of other layers. Technologies such as GMPLS have been proposed for a number of years to achieve convergence but have never been deployed across layers in operational networks. Control scaling, complexity, interoperability, and organizational fit have been issues. This is a key area of focus for ONF in providing specifications and technologies for integrated control across layers.
There are many reasons for doing this with some potentially important use cases including:
- Expressing of traffic at the lowest layer possible, to reduce the cost of transmission;
- Automatic turn-up of new express links at lower layers to alleviate congestion at higher layers;
- Periodic re-optimization of service paths, to address service churn, connection revisions and addition of new capacity;
- Coordinated protection and restoration to achieve optimum performance versus the cost of each layer, based on the traffic demands.
Coordinated restoration has been an aspiration of many network operators for many years. Typical use cases include protection of high priority traffic using higher layer mechanisms such as MPLS fast reroute or OTN SNCP, with restoration of lower priority traffic using wavelength reroute – a slower but very low cost approach.
So what do the networks need to make this a reality? You can find out in the full article featured in Global Telecoms Business on page 64by clicking here.
- Lyndon Ong, Principal Network Architect and Technical Fellow at Ciena, and ONF Optical Transport Working Group Chair; and Paul Littlewood Principal Network Architect at Ciena
