Network virtualization is such a hot topic that all developments, announcements, suggestions and proposed standards...
threaten to muddy the big picture. Operators have long planning cycles, and they have to understand where the virtualization trends will take them down the line, even as far as 10 years out. So, what will virtualized networks look in the future?
The basic notion of a virtual network is it behaves like a real, dedicated network, but it is in fact created by shared technologies, not dedicated devices. Virtual networks can be created in four ways, and the interplay of these approaches will, in turn, create the future of virtualized networks.
Segmented switching and routing approach
The first approach uses segmented switching and routing, which is how virtualized networks have traditionally been built. If we assume virtual networks of the future are built only to generate virtual private network (VPN) or virtual LAN (VLAN) services, this model would surely dominate 10 years from now.
This segmented approach will contribute less to virtual networking over time, however, and it will be relatively unimportant in a decade. The reason is most virtual networks are built to support multi-tenant cloud computing, and segmenting real devices to create cloud networks doesn't scale or provide as much multi-tenant independence as operators and enterprises need.
The biggest problem in extending segmented hardware infrastructure virtualization into the future is the sheer scale of cloud virtual networking. Standards like virtual extensible LAN expand the limited number of virtual networks VLANs can support. And while MPLS VPNs have essentially unlimited scope, they are expensive for operators and users alike.
In addition, making constant changes to network routers and switches for tenant services includes the risk of accidentally creating instability and affecting other customers. Network services based on infrastructure virtualization usually require special hardware, which means they're difficult to deploy in the public cloud.
Overlay network approach
Overlay networks, which include software-defined WANs (SD-WAN), are the second model to create virtualized networks, emerged directly from cloud computing. Early cloud providers wanted easy multi-tenant networking, and Nicira -- now part of Dell Technologies and VMware -- provided the first popular implementation. Other vendors, like Nokia's Nuage Networks, also support the overlay approach. Most recently, the SD-WAN approach has taken overlay networks to the heart of the segmented-infrastructure opportunity, creating VLANs and VPN services. The key in the overlay model is to create a network above a network for each tenant.
A number of overlay strategies are available, and standardizing on only one overlay approach isn't necessary, since a number of approaches to virtual networks can work. If the approach supports software clients, for example, overlay virtualization can easily be extended to cloud applications.
This virtualization model is based on edge technology, not on changing the network. As a result, it can be applied by users, managed service providers and network operators. This broad usability by different customer groups means a lot of market support and competition among both SD-WAN vendors and SD-WAN providers.
Virtual switches and routers
The third virtualization model is to deploy virtual switches and routers hosted on servers and connected via tunnels or virtual wires. Those virtual wires are provided by traditional switching and routing, including MPLS. The idea of cloud-hosted network elements comes from network functions virtualization (NFV).
So far, however, NFV hasn't explicitly targeted switching and routing, focusing instead on functions like firewalls and virtual customer premises equipment. But virtual switches are already widely used in cloud computing, and using virtual routers as gateway devices in the cloud is gaining traction.
The challenge for virtual switches and routers is in deploying them outside the data center. WANs can be built using virtual switches and routers, but those tunnels or virtual wires aren't offered widely as network operator services. Without tunnels or virtual wires, a virtual router or switch ends up looking like an SD-WAN device and an internet overlay. If virtual wire services were widely offered, this virtualization model could perhaps replace traditional VLANs and VPNs. But without those virtual wire services, the future of virtual switches and routers is murky.
Software-defined networking (SDN) is the final virtualization option that can replace the traditional adaptive model of route management in both switches and routers with explicit, centrally managed forwarding via white box hardware or software switches and routers. OpenFlow, the protocol used to control forwarding in SDN, has gained traction in the data center, but less so in the WAN. The problem may lie in making central forwarding control scalable or in fighting off the lower-cost approach offered by SD-WAN.
One mission for SDN could be supplying those virtual wires. Unlike internet connectivity, which is massive and changes dynamically as users click on URLs, virtual wires would likely be fairly persistent and pose few scalability problems. If network operators were to start providing virtual wires via SDN, they could then overlay SD-WAN or virtual switches and routers on top of the network, thereby creating a new model for many services.
SD-WAN paving the way
So, where do we end up envisioning the path to virtualized networks of the future? SDN and NFV as virtual network drivers are both dependent on major shifts in how network operators build their infrastructure, which limits how fast they can be adopted. Both are also in search of a new services model to generate new revenues. Without SD-WAN, you'd have to assume currently noncooperative approaches -- SDN, NFV and infrastructure virtualization -- would somehow converge on a common strategy to drive change.
As the virtualized network strategy with the most going for it, SD-WAN has the most fundamental role in virtualization and the broadest base of interested parties, not SDN or NFV. SD-WAN could be the transformational strategy of the current age. Not only does it disconnect service from the details of infrastructure, SD-WAN opens the door for infrastructure to change by adopting other virtualization technologies, including SDN and NFV. Because SD-WAN runs above infrastructure, and because it can be deployed by enterprises, network operators and managed services providers, fear of change or high levels of legacy infrastructure cost can't stall it out.
Ten years from now, most business services probably will be SD-WAN-based, and cloud service delivery will be dominated by SD-WAN, too. Over the next decade, SD-WAN will both pave the way to change and define how network changes are matched to consumer and business services. The bottom line is SD-WAN will get us to the viable virtualized networks of the future for operators and enterprises alike.
How SDN could benefit from virtual wires
What to expect from SD-WAN in the more immediate future