Networks for a yottabyte world

The future requires networks that are qualitatively more flexible, scalable, efficient and manageable. Network function virtualization and software defined networks are the way forward

The rate at which telecom networks are growing and changing is nothing short of fantastic. It’s always risky to embrace a new paradigm, but for network carriers and customers, the risks of waiting could be greater.

Network function virtualization (NFV) and software defined networks (SDN) represent a radical departure from the traditional way of building, managing and evolving telecom networks. It’s often described as a switch from proprietary boxes to commercial-off-the-shelf (COTS) hardware. While there is a potentially significant cost-savings in making such a switch, cost-savings is not the main driving force. The ability to quickly implement new business models, to deliver applications on demand, and to automatically provision and tear down resources are what make NFV and SDN so potentially disruptive.

Not your father’s networks

The growth in telecom traffic over the last 25 years has been mind-boggling. Since introduction of the Mosaic web browser in 1993, Internet traffic has grown by a factor of 10 million. Since the introduction of the iPhone in 2007, mobile data traffic has grown by a factor of 1,000.

While these numbers partly reflect humble beginnings, growth in Internet and mobile IP traffic continues at an exponential pace. According to Cisco, over the next year monthly Internet traffic is expected to increase by approximately 24,000 petabytes. Monthly mobile data traffic is expected to increase by nearly 8,000 petabytes. To put this into perspective, consider that it would take 2,000 years to play one petabyte of MP3 music.

The growth in network transmission rates has also been amazing. In 2000, cable modems offered a modest performance advantage over dial-up modems. Today, cable operators are delivering speeds up to 1 Gbps. Mobile networks, on the other hand, had no speed advantage over dial-up modems in 2000. Recently, 4G networks have demonstrated the ability to deliver 1 Gbps, while emerging 5G networks promise peak speeds up to 20 Gbps.

Latency can be just as important—or even more important—than data rates. In highly interactive applications, fast data rates can’t make up for slow response time. This is why 5G networks promise latency on the order of 1 millisecond where needed. (You can’t get much better than that without bumping into the laws of physics.) Low latency requires bringing the applications and data closer to users—edge computing.

There is a parallel effort to digitally transform operators’ OSS/BSS (operations support system/business support system). These are applications that sit on top of the network. In the past, it took operators months of planning, development, and testing to introduce new services for customers. That is no longer acceptable—competition waits for no one.

NFV/SDN vendors and early adopters

Where do we stand today? Most of the people I’ve talked to say that the evolved packet core (EPC) has been largely virtualized. Beyond that, NFV adoption has been slow, and NFV projects often fail to achieve their goals. But it’s still early.

SDN is often described as separating out the control plane from the data plane in the boxes that comprise a network. Kumar Mehta, Founder and Chief Digital Officer at Versa Networks, also sees value in separating out the analytics plane. Versa’s software platform uses analytics to provide better user experiences, more efficient use of network resources, and security. Using analytics and artificial intelligence, features such as forward error correction and packet replication can be activated automatically, and traffic can be assigned to the predicted best paths or circuits.

Mehta says that NFV/SDN growing pains are inevitable. For instance, virtualization prevents vendor lock-in, but when there are multiple vendors and a problem appears, vendors tend to point fingers. The industry is working on ways to prevent finger-pointing, and Mehta believes it will become less common as NFV/SDN matures.

Blue Planet, a division of Ciena, sees 5G as a major opportunity for NFV/SDN. While 4G networks were built on a traditional best-effort basis, 5G networks will be more flexible and dynamic—they will be capable of being reconfigured and optimized on-the-fly. They will also be more automated, requiring less manual effort to provision resources and make the most efficient use of bandwidth. It will take time to implement NFV/SDN, requiring personnel with new skills and a new operator culture.

Blue Planet’s products include MDSO (Multidomain Service Orchestration for multi-vendor networks), Blue Planet MCP (Manage, Control and Plan for Ciena networks), and Blue Planet Analytics.

F5 Networks describes itself as delivering “cloud and security solutions that enable organizations to embrace the application infrastructure they choose without sacrificing speed and control.” The company sees operators choosing between two roads to NFV/SDN: wholesale migration or incremental (project-based) change. Most operators are choosing the incremental approach. AT&T is an exception with the bold vision outlined in its Domain 2.0 initiative.

F5 focuses on making sure that application services such as firewall, video optimization, and parental controls are always available. To that end James Thomson, Solution Architect, sees Tier 1 mobile operators increasing the number of their data center locations by a factor of 10. That will require the ability to spin up new resources in minutes or hours—the old ways took months because they were too labor-intensive.

Legacy vendors are also getting on the NFV/SDN bandwagon. Swedish telecom giant Ericsson today describes itself as a provider of “Information and Communication Technology (ICT) to service providers.” Ericsson points out that 5G will serve a wider range of applications, requiring more flexible networks with more local resources. In some cases, edge logic will be deployed in small cells or even on the customer’s premises. Ericsson is a big proponent of the Internet of Things, but recognizes that sending data to the cloud doesn’t always make sense. Autonomous vehicles can’t wait for decisions from remote sites. Enterprise customers may rent space on virtual machines and install their own software—rather than deploying large amounts of hardware as in the past.

While the goal is to virtualize as much of the network as possible, some elements (such as antennas) can’t be virtualized, while in other cases virtualization is possible but hardware solutions remain more efficient.

Since 2014, Ericsson has amassed 100 customers for its virtual evolved packet core (vEPC). Operator customers include Softbank, Swisscom, Telstra and Vodafone. Some of the first projects focused on voice over LTE (VoLTE) and Wi-Fi calling.

As described in its visionary Domain 2.0 white paper, AT&T sees the need for a new architecture inspired by cloud technologies enabling new business models, greater customer value, and a larger and more diverse choice of suppliers.

Chris Rice, Senior VP of AT&T Labs, believes that going forward operators will benefit from a new ecosystem, an open source approach to network solutions, and a devops culture. AT&T’s goal is to implement NFV/SDN in 75 percent of its network by year 2020. In Rice’s opinion, the problem with the incremental approach is that it’s too easy to lose sight of the larger goals.

Rice is also bullish on the role of AI in enabling network automation. Because AI is still evolving, it is currently best implemented in an “open loop” configuration, providing recommendations to network personnel. As AI becomes more reliable, operators will migrate to a “closed loop” model in which actions are taken automatically.

When change is the only constant

NFV and SDN are inevitable developments. While there will be much turbulence ahead for early adopters, and some proprietary solutions will always be with us, once NFV and SDN reach critical mass the advantages will be overwhelming. At the rate that information is being generated, collected and mined, the network industry will have no choice but to go to a more flexible and scalable approach.

This post is based on commentary by Ira Brodsky that first appeared at Computerworld. Brodsky is a Senior Analyst with Datacomm Research and is the author of five books about technology. Brodsky focuses on mobile solutions for payments, retail automation, and health care.