Sales of software-defined networking (SDN) gear and software might well reach $35 billion by 2018, but sales are likely to be about $3.4 billion in 2014, according to SDN Central.
That illustrates a principle about big changes in communications network architecture: such changes tend to take longer than expected, and that is likely to be the case for some parts of the enterprise or data center SDN and telecom carrier network-function virtualization (NFV) architectures that promise widespread use of open systems and standards-based components.
As always, with any major change of networking architecture, there are many practical implementation issues.
As with any “open” approach, it will take some time for full interoperability across device and software domains to occur, even if the basic principle is simple. Typically, such transitions involve hybrid approaches, at first.
Since there are so far no widespread standards for integration, suppliers have to create and test on a bilateral basis, for firewalls, load balancers and switches, for example.
In most cases, where a “rip and replace” approach is not preferred, suppliers also will implement overlay and underlay solutions, encapsulating existing functions. That takes more time than simply implementing to a universal standard.
Also, it remains a bit unclear whether just one framework, or possibly several, are emerging. SDN is an architecture that makes it easier to build and efficiently manage large complex networks. SDN was developed for enterprise and campus communications.
Even there, early on, many would have used the term SDN synonymously with OpenFlow, a communications protocol.
But a related concept - NFV - also was created to increase network resource use, from a communications service provider orientation.
NFV is an initiative driven by the ETSI Industry Specification Group to virtualize network functions previously performed by proprietary dedicated hardware.
So SDN and NFV overlap. SDN focuses principally on “virtualizing” by separating control plane and data plane functions. But so does NFV.
Some might argue both are complementary. That might be hard to discern.
SDN might be said to provide network flexibility by combining centralized control functions and widely-distributed data plane functions. The value for a network operator would be an ability to evolve and innovate faster, as well as lower both capital and operating expense.
But NFV also is intended to reduce network operator capital and operating expense by allowing use of lower-cost network elements and reducing power consumption.
As a practical matter, many observers would say SDN and NFV overlap to a high degree. To date, some might say SDN has gained traction in data center and cloud computing environments, while telecom carriers are driving many of the NFV efforts.
There are several key areas of overlap. Both SDN and NFV emphasize use of “virtual” resources rather than physical resources. Both imply use of simpler, general purpose network elements rather than dedicated network elements.
Likewise, both SDN and SDN support network services (load balancing, traffic analysis, monitoring) that are managed and implemented on a virtual basis, rather than in dedicated appliances.
Also, both SDN and NFV might be said to support virtualized applications that are cloud based rather than implemented locally.
So some might argue the differences are “intentions.” SDN was created to abstract network functions, enabling faster innovation, while NFV was created to reduce capex, opex, reduce power consumption and save space.
For communications service providers, as opposed to data center and cloud services operators, there is an important related concept, namely separation of application logic and enforcement from corresponding subscriber data, roughly corresponding to the separation of control plane from data plane.
Telecom guys have experience with distributed Control Plane (Management Systems, PNNI from ATM) while they are less comfortable with virtualization. The data center expertise is vice versa. Hence, two worlds meet and it is well understand from both sides, that their competence will be of benefit for the other “world”. Such cross-fertilization in the past created significant steps in network innovation. There is good reason to have high hopes…[this might be compared with communication meeting computers -> which resulted in the Internet]
As always, data center professionals might know why they like SDN, as telecom service providers likewise know the advantages of NFV. But “flash cut” approaches will not be possible in most cases, meaning adoption will move in phases.