Software-defined networks (SDNs) are at that sensitive point faced by any emerging technology. On one hand, the essential concept is fairly straightforward. SDNs enable far more fluid, fast and efficient networking by facilitating communications between management software from all network devices – regardless of vendor – and centralized, intelligent decision making on how to use routers, switches and other network elements.
A couple of weeks ago, Transworld Data president Mary Shacklett, writing at ZDNet, provided a great overview of SDNs. The piece presented 10 data points about the technology, the second of which provided a good illustration of what SDNs do:
There are two planes in network devices — a control plane that determines where traffic is sent and a data plane that forwards traffic based on what the control plane tells it to do. With SDN, these two planes have been detached (or decoupled) from each other. The data plane (or data forwarding plane) remains with the network hardware — but the control plane (or controller) that makes decisions about where traffic will be sent is now executed through software. This separation makes network virtualization possible because you're no longer executing all the command or control rules on the hardware itself.
Lots of new technologies make sense. Often, however, the needle they have to thread is as much business as technology oriented. At eWeek, Jeffrey Burt discusses where SDNs are on the hype cycle.
The norm is happening: Vendors are using the term loosely so they are not perceived as laggards, even if they don’t have true SDN equipment. Few real-world deployments are evident, though big vendors claim to have many ongoing tests. However, he concludes that the fire of real and meaningful advancement is shrouded within by the smoke of press releases and vaporware:
However, that's not to say SDN is all hype. The technology holds the promise of solving some significant problems in the data center around issues of networking complexity, cost programmability, flexibility and scalability. So both vendors and analysts are expecting the number of SDN test drives to grow as 2013 wears on. Real deployments of SDN systems are expected to start taking off late this year and early next.
The point is that this would be monstrously complex stuff if it was being built into new systems without legacy infrastructure. Bringing it to market in an environment where various network elements can’t retain their book value and can’t be easily changed increases the level of difficulty. The backdrop to all this is that the technical community hasn’t fully defined SDNs.
SDNs will succeed because they offer real value. The payoff is speed to market, according to Bethany Mayer, the senior vice president and general manager of HP Networking. During an interview at HP Discover 2013, Mayer said that SDNs enable changes to be made in a far shorter timeframe than in an environment in which every network element must be manually adjusted. SDNs do not add capacity to the network. Instead, they make what is already there available more fluidly, according to the report at SiliconANGLE.
As if that all isn’t enough, the industry should get ready for the next wave of hype. At least two related concepts – network functions virtualization (NFV) and network virtualization (NV) – are on the horizon. The first refers to running demanding network functions, such as firewalls, in software on virtual machines and commodity servers. Network virtualization focuses on decoupling virtual and physical resources to provide more flexibility. At the highest level, SDN, NFV, NV and other technologies deal with enabling network operators fluidly and seamlessly tap into a pool of resources.