Telecom’s Twin Peaks

Software-Defined and Virtualized Networks

By Christian BartoschMichael HitzRüdiger SchichtMaikel Wilms, and Michael Sherman

Software-defined networking (SDN) and network function virtualization (NFV) have already started to change the way that telecom networks are designed, deployed, and operated. Together, these two technologies can not only reduce networks’ capital and operating costs but also improve their flexibility and scalability. Moreover, their open-source architecture can foster greater agility and innovation.

But will operators strike gold with these technologies? Or will the biggest winners be today’s sellers of pickaxes and shovels, the vendors of SDN and NFV?

We believe that telecom operators stand to gain real benefits from SDN and NFV—but achieving those benefits will be challenging. Operators will need to fundamentally revise their operating model, skill profiles, and organizational structure to make way for software-enabled and virtualized networks.

It’s important that operators move swiftly toward this new world, in which network functions will no longer require specialized and dedicated hardware—but they should not do so at the expense of preparation and foresight. Rather than rush in and replace traditional network elements that have many years of useful life left, operators should develop a careful and comprehensive transformation that goes well beyond network technology. To ensure that the transformation stays on track, the CEO will need to champion and oversee it.

The Promise of Software and Virtualization

Google’s introduction of a global SDN backbone, called B4, helped usher in the largest shift in networking since packet communication. SDN allows network administrators to use software to manage traffic by deploying large numbers of inexpensive switches running in parallel. SDN essentially separates a network’s brains (or software) from its muscle (or hardware). NFV, for its part, allows engineers to replace traditional network devices, such as routers, with software running on commodity servers.

In both cases, open-source software replaces many of the functions traditionally performed by dedicated, complex systems. SDN and NFV are two sides of the same coin; we refer to their collective use as software-defined virtualized networking (SVN).

With traffic growth exploding, SVN deployment has become a key lever for managing costs. It can already accommodate data throughput sufficient for all but the biggest network traffic challenges. Early adopters are suggesting that they will likely save 30% to 50% on capital and operating costs over five years. One example of the enthusiasm over SDN’s potential: AT&T aims to have 75% of its networks virtualized by 2020.

The Transformation Challenge

The shift to SVN will have significant implications for network architecture. SVN will require a high degree of automation and complexity, as software mediates the relationships among network nodes, customers, and services. Self-organizing network functionality and smart agents will play critical roles. SVN, in other words, essentially combines the complexities of both networks and IT into a single challenge.

Many of the concepts underlying SVN have been around for years. Software and virtualization have already started to replace certain network elements, such as mobile switches. But the technology is in flux. The target state for the architecture and operating model is not yet fully understood, let alone tested and proven. Operators such as AT&T and Verizon have published proposals for reference architecture in cooperation with major vendors. But we expect that Google, the Open Networking Foundation, and Silicon Valley startups will be the leaders in fleshing out the SVN concepts.

One of the largest challenges will be the integration of SVN solutions into existing network infrastructure and operational processes. Given the radical departure from traditional network upgrades, SVN deployments’ integration costs generally make up a greater share of overall cost than is the case in more traditional settings.

Because of these uncertainties, operators must be willing to adopt a test-and-learn and agile approach to transformation. Otherwise, they will be unable to change course when the technology shifts.

Several attributes will enable a successful transformation:

  • Virtual development and operations teams that work seamlessly across organizational boundaries
  • A strong open-source-architecture team consisting of both internal and external people
  • A joint commercial and technical team to migrate customers systematically to new offerings

In addition to transforming their technology architecture, operators will need to make changes to their organizational structure, capabilities, and behavior. As part of its 2020 virtualization effort, for example, AT&T is investing in building fundamental new skills among its engineering staff.

Operators must choose their partners in transformation carefully. Large network vendors, for example, have already been protecting their legacy portfolio by attempting to sell SVNs at a premium over legacy applications. Others may be developing customized solutions that are misaligned with evolving industry standards. Switching vendors can be difficult, but it’s often necessary.

A Pragmatic Approach

Initially, operators should focus their virtualization efforts on those parts of the network that will bring fast and tangible benefits without introducing significant business risks. While the sequencing will differ by operator, some general patterns are emerging. Many current approaches are based on replacing current hardware platforms by porting their respective software to a virtualized environment running on commercial, off-the-shelf servers. A key challenge is how to operate these virtualized functions in a cloud environment. The required management and orchestration tools are still in development.

Mobile Networks. SVN is gaining the greatest traction in services rather than in the access and core domains of mobile network operators. This is not surprising. An SVN-based architecture can address the scale and cost issues that led data center engineers to consider software and virtualization options in the first place. Today, some virtualized core functions are almost routine. Virtualized VoIP controllers, for example, have already been deployed in more than 100 mobile networks, so the performance gains and other benefits are well known. Virtualized packet cores are the next area of deployment.

Given the relatively static nature of routing in the backhaul and transport of mobile traffic, operators are not yet relying heavily on SVN in their legacy networks. In the core domain, latency also can become a major bottleneck for SVN deployment. The speed of light limits the distance between switches and the SVN controller to less than 300 kilometers. To take advantage of SVN in larger networks, operators would need to subdivide the control architecture, introducing additional complexity. At longer distances between switches and controller, the benefits of SVN are limited to an advanced management system that centralizes provisioning and facilitates automation.

In the wireless space, SVN does make sense in new, large-scale greenfield or capacity-expansion deployments in public spaces such as shopping centers and airports. The bandwidth needs mandate a large number of radio cells, and the performance of such an ultrafine radio mesh depends on efficient communication among the respective baseband units. It is actually easier to configure these units as applications running in parallel on a cloud than as discrete systems communicating via physical wires. Likewise, SVN can help tie together radio networks of different cell types and access technologies. One example would be to dynamically balance loads in large, distributed antenna systems.

Fixed Networks. In fixed networks, unlike in mobile networks, SVN’s greatest immediate potential is in the access domain because of fixed networks’ more dynamic routing, which requires updates to routing tables and firewall policies. This sort of environment plays to SVN’s strengths because it allows the use of inexpensive and remotely managed switches on customer premises. Functions like firewalls and border gateway routing can be conveniently moved to an edge router managed by an SVN controller, reducing the frequency of configuration mistakes and technician visits to customer premises.

SVN’s greatest potential will vary by segment. In the B2B market, both SVN overlays on top of existing routers and pure deployments of SVN equipment make sense. In the B2C market, SVN holds great promise for cost reduction and operational simplification, too. For example, Bayonette’s vHome offering is based on this approach.

Converged Networks. The combination of fixed- and mobile-service platforms offers a strong rationale to migrate to SVN. Many integrated operators have introduced or plan to introduce voice over LTE or voice over Wi-Fi. One operator has created a converged network across fixed, cable, and mobile by implementing virtualized IP-based solutions for multimedia, voice, and data. This approach significantly lowered capital and operating costs and reduced the number of distinct platforms from six to two.

A Top Priority

An SVN transformation will last several years and come with great rewards and risks. Operators that act swiftly can build competitive advantage in both agility and cost structure. Slow-footed operators, on the other hand, risk becoming a “dumb pipe” or simply unable to compete.

To win, operators will need to devote resources, manage risks, and act decisively. SVN needs to be a top priority for CEOs. If it is not, the promise of transformation will dissolve into a puddle of piecemeal and incremental improvements at best.