David Wilde, AARNet’s Network Architect weighs in on the hype, reality and potential of Software-defined Networking (SDN)
A presenter at the recent Internet2 Global Summit in Denver noted that until recently, “Cloud” was the buzzword of choice – breathtaking in its lack of clear definition and unlimited in its ability to be applied to any IT project. But Software-Defined Networking or “SDN”– a concept hyped to change the world of IT and networking, as we know it, has now knocked Cloud off this lofty pedestal.
SDN is appealing because of its potential for significantly reducing operational expenses. It enables administrators to reconfigure networks of switches and routers easily and quickly on demand in response to business needs from a central controller, without having to manually reconfigure hardware.
Traditional network devices, such as the switches and routers, which make up most enterprise networks, perform two basic tasks: they run a control plane and a data plane.
The control plane is the brains of the operation: communicating with neighbouring network devices, deciding on the best path to forward traffic, sending and receiving management information. This is implemented by the operating system of the network device, and network vendors such as Cisco, Juniper, HP and others have invested enormous amounts of time and money in the features and functionalities of their particular operating system.
The data plane is where network traffic is forwarded: typically by the specialised hardware built into the ports and backplane of the device.
The SDN model simply decouples these two planes.
The control plane, being essentially a software activity, can be shifted out of individual network devices and onto either centralised or distributed compute platforms.
And the data plane, being essentially a hardware activity, remains the task of the network device.
The SDN model opens up new possibilities, such as:
SDN isn’t applicable to every environment.
SDN implementations are addressing some technical and business requirements today.
This environment is most suited to the SDN model: a dense deployment of switches; rich and reliable interconnectivity to ensure the newly decoupled control plane doesn’t lose contact with the devices it’s meant to be controlling; and typically a single organisation controlling the network, the devices connecting to it and the services they’re running.
Vendors are releasing centralised controllers where an administrator can define user and group policies. In this way, the lower level device configuration is abstracted and hidden away from the network administrator, allowing them to focus on translating business requirements into network policy.
RMIT University is taking an even more ambitious approach, deploying OpenFlow-capable switches across its entire network, which covers hundreds of buildings in Melbourne’s CBD as well as several remote campuses.
(“OpenFlow”, another heavily hyped buzzword, is a protocol defining how the servers running the control plane communicate with the switches running the control plane.)
In a promising collaborative approach between the RMIT ITS team and Electrical and Computer Engineering researchers, a network-wide controller application is being developed to migrate entirely away from the traditional device-by-device network model and towards a centralised controller interface architecture.
Moving from the data centre and campus out into the WAN (Wide Area Network), there is an entire sub-domain of SDN known as NFV: Network Function Virtualisation. In the same way that virtual machines have revolutionised the world of servers, NFV has the potential to open up wide area networks to segmentation and virtualisation, extending the capabilities of today’s MPLS-based VPNs to include the programmability of SDN.
Will carriers and service providers choose to move away from their traditional vendors? Too early to judge at this stage.
SDN is still very much a burgeoning area. Researchers across many Australian universities are deeply engaged in exploring and exploiting the possibilities. Some current areas of research include security, access control, bandwidth utilisation & optimisation, traffic differentiation & QoS, cost reduction and integration with the NBN.
One of the attractive aspects of SDN is its ability to level the playing field. Expensive instruments or resources generally aren’t required, so an SDN researcher in Australia has access to the same resources as a researcher in the USA or Europe. AARNet aims to enable Australian researchers to reach the forefront of this international stage and is involved in two new SDN initiatives:
Feedback and more information to David Wilde (Network Architect)
Oct 13, 2017