The Software-Defined Internet

The Internet continues to change. The traditional Internet was a highly static system: All addresses were global, endpoints were stationary, users were tied to a single primary device, and the application space was dominated by the Web. The current Internet has already escaped that definition. Today, devices switch seamlessly between IPv4 and IPv6; devices are mobile, switching addresses as they roam; users have multiple devices; and applications are more diverse than ever. The future Internet will feature more devices than people, a more robust security architecture, and a pool of services that matches the diversity of applications. The boundary between processing and communication is already fuzzy, as we see when we examine the services offered to residential and commercial Internet users. It will soon be impossible to draw a coherent line between processing and transmission,  if it isn’t already.

One way to describe this trajectory in simple terms is to say an Internet defined primarily by hardware is giving way to one defined by software. Hardware is devices, switches, and locations, but software is whatever we need it to be. One device that undergo a complete transformation is the home router or residential gateway.

Gateways are sad little devices today, mainly produced by small, low-margin companies with limited engineering resources. These companies assemble components built by larger chip companies who supply the gateway firms with circuit diagrams and bundles of software. These reference designs, as they’re called, or either lightly modified or not modified at all. In many cases, the firms who manufacture and sell gateways don’t even understand all the parts and functions they contain. I can say this because I was once in the home router business.

The economics of the gateway business don’t lend themselves to major upgrades. Consumers rarely update software, even though all the software in these devices is out-of-date and buggy. While techies are happy to control their home gateways, the average consumer is simply annoyed by ongoing maintenance tasks and the endless tweaking of obscure parameters.

When we view the home gateway as a software system that provides the first and last point of entry to and from the Internet, we begin to reformulate these devices as edges of the Internet service. It’s not surprising that service providers see gateways that way, because they’re the parties most affected by the shortcomings in today’s devices. Hence, companies like Ericsson and Alcatel-Lucent who cater to carriers are promoting the vision of the “virtual home gateway,” a software system that joins customer premise equipment to software-defined network switches within the carrier footprint. Ericsson’s white paper Virtual CPE and Software Defined Networking articulates this vision, as does Alcatel-Lucent’s Virtualized Residential Gateway.

From the consumer perspective, it’s beneficial to outsource the management of our home networks to experts. As we add our wearables, fitness devices, security cameras, smart home devices (lights, thermostats, door locks, geofencing, etc.) to our Wi-Fi networks, it’s good to have someone minding the interactions and updates these devices entail. This could be done by OTT service providers, but ISPs are well-positioned to move into this space, as are home security firms.

The transition of the Internet from a device-dominated system to a software defined one is potential more important and more hazard-prone than the transition of the old telephone network into an all-IP Internet tributary. One implication of the SDN revolution is new service models.

The FCC’s Technical Advisory Committee (TAC) is more concerned than ever with the tension between the traditional Internet and the Next Generation Internet. At the December 6th meeting, the TAC heard about the IoT, SDNs, and virtual networks from half of its six major working groups, Cybersecurity, NG Internet Services, and Future Game Changing Technologies.

The major takeaways were security – which has to be done everywhere – and differentiated services. There’s no doubt that the TAC marches to the beat of a different drummer than the 8th floor managers who devised the Open Internet order, which bans “paid priortization;” the NG Internet Service group says: “No differentiation without remuneration” and stresses the need to measure Internet Quality of Service and Quality of Experience.

NGIS is right, of course: An exploding universe of applications cries out for service differentiation, but differentiation calls out for a Quality of Service floor that prevents the collateral damage envisioned by some advocates when for-fee QoS in on the table. A software-defined Internet can do both, even if it’s not clear that the traditional QoS mechanisms, IntServ and DiffServe, can. While I think the traditional mechanisms are up to the task, they can be hard to manage, hence the added power of SDN’s ability to increase bandwidth on the fly is very beneficial. An SDN system can deal with QoS with bandwidth-on-demand, a tool that’s outside the traditional Internet’s arsenal. An SDN Internet can also limit security exposures by segregating trusted services from unknown ones.

With SDN, we can easily migrate from TCP/IP toward more advanced networking technologies such as RINA, which I discussed with its inventor John Day in a recent podcast. It’s actually possible to convert an old-fashioned home gateway into an SDN device controlled by OpenFlow, a tool used in industrial networks; the paper Network Functions Virtualization in Home Networks shows how.

Even though Washington does its best to dash our hopes for a better Internet, the technology hasn’t stopped moving. So enjoy the holidays and check in after Christmas for some more optimism.