The networking industry often has a problem taking the real measure of new technology, so potential changes can...
be both over- or underestimated. Either of these mistakes can throw off the network operators' planning. For example, software-defined networking and network functions virtualization have been overhyped in terms of some uses and underestimated in others.
So, where are these two technologies really going? A good place to start forecasting is in the carrier cloud, which may drive the use of SDN and NFV technologies more than anyone expected a few years ago.
The original concepts of software-defined networks and NFV were to replace some existing network devices with new technology. SDN was to substitute white box switching for current Ethernet and IP devices, while NFV was to replace custom appliances like firewalls with hosted instances. The original vision was this combination would radically reduce Capex. Over time, however, it became clear the effects of reduced Capex alone wouldn't justify either technology.
The current focus is on improving Opex and increasing operator competition in a market increasingly dominated by over-the-top content delivery. But the effects of SDN and NFV were overestimated in their ability to decrease Opex, too. SDN and NFV effectively replace existing devices, so there is no reason to think a one-to-one replacement of a router with a virtual or SDN router would reduce operations costs.
In fact, in existing networks, SDN's central management can replace adaptive management -- the use of automatic actions among devices that include automatic rerouting around an IP failure, rather than requiring external management commands. But that use could increase Opex. In NFV's hosted instances, in other words, each virtual function that is separately deployed requires the coordination of cloud deployment and connection, which is also likely to increase costs.
The reality is both SDN and NFV will require service lifecycle management, in addition to the new technologies, to absorb additional management complexity and to bring about the improvements in agility and operations efficiency that operators want. Without service lifecycle management, the complexity inherent in SDN and NFV technologies would make them far less practical. But whether an organization needs SDN and NFV in addition to service lifecycle management depends on the other value propositions the technologies offer, since most of the advantages are really benefits of service lifecycle management and, therefore, are available without SDN or NFV.
Because an optimum implementation of a new service lifecycle automation layer would also improve Opex and agility for existing networks, the question concerning the future of SDN and NFV technologies is how they could be used to improve existing networks further or used in new areas to create new services, rather than hosting versions of older services.
OpenFlow SDN in the carrier cloud and beyond
The broadest effect of SDN on the future network will be more implemental than revolutionary. Software can define a network in many ways, not just by replacing routers with white box OpenFlow switches. In fact, most existing SDN use by operators doesn't involve OpenFlow, but rather depends on policy distribution and enforcement to control network behavior. This is likely to continue to be the largest SDN mission for years to come.
But pure OpenFlow SDN has already earned a place in many cloud data centers, because SDN allows full control over data paths, the meshing of devices and multi-tenant service separation. As network operators deploy carrier cloud technology for any service mission, they'll likely increase their SDN use proportionally. In the future, SDN can be expected to be the basis for nearly all operator data center switching -- both for cloud and virtualized data centers, and even for bare-metal servers.
OpenFlow SDN's use outside the data center is likely to be less the replacement of switches and routers and more in creating electrical-layer virtual wires above agile optics. SDN forwarding has low overhead and latency, and using it to create separate paths from fat optical pipes would introduce minimal additional delay. Virtual wires could then be combined with customer premises equipment (CPE) to create a software-defined WAN, or with software routers to create virtual private networks.
NFV's place in the networks of the future
NFV is currently used primarily to deploy virtual edge features like firewalls onto general-purpose devices at the service demarcation point, which is called the virtual CPE (vCPE) model. In the future, NFV clearly will be a technology for deploying the edge functionality of SD-WAN, and for deploying software router instances that link SDN virtual wires. These two applications are directly linked to business service users, rather than residential users, because residential vCPE applications would have to compete with low-cost devices like broadband gateway hubs with fairly static features.
Virtual CPE is a customer- or tenant-specific application. Another pathway to new NFV applications is to replace network devices that provide multiuser services with software. This is already happening with mobile infrastructure, where some IP Multimedia Subsystem and Evolved Packet Core components have been converted to software images.
Content delivery network components could also be dynamically deployed using NFV. As the 5G wireless broadband standards advance and begin to drive deployment, some next-generation elements of a radio access network and features to support network slicing may also be deployed by NFV.
The most exciting new set of NFV applications comes from the personalization and contextualization of consumer services, including advertising and video, and from the internet of things (IoT). Personalizing services involves the use of an agent process that collects information on behalf of the user. Advertising and video personalization and contextualization would tune information based on the geographic or social context of the consumer.
Likewise, IoT is a maze of event distribution and event-processing elements. All the pieces of these applications could be deployed using NFV, but alternative approaches based on microservices and functional programming -- like Amazon Web Services Lambda -- might be used, as well. NFV's final role here is still up in the air.
On to the future, into the WAN
In the final analysis, the roles for SDN and NFV technologies may be determined by the carrier cloud, rather than these technologies driving the carrier cloud. With more cloud data centers, the more SDN will be deployed in those data centers and the more likely it will be to extend into the WAN. For example, building data center interconnects using SDN can be done more easily if SDN is already deployed in the data centers.
As more cloud data centers are deployed for hosting functions, the more viable NFV's economics become. If neither SDN nor NFV can control their own destinies, the industry will have to start thinking a lot more about the role of the carrier cloud and about general trends in cloud computing, because in the end, SDN and NFV may become more powerful by being incorporated in broader cloud evolution.
Virtual wires could get SDN out of the data center
What are the two main reasons that service providers deploy NFV and SDN technologies at different paces?
Moving SDN from the data center could help broader network initiatives