Software-Defined Networks Explained

This new approach might be the future of computer networks

7 August 2013

Software-defined networks are being touted as a disruptive technology that could revolutionize computer networks through software, not hardware. But little is known about the architecture that will put the capabilities of running such systems into the hands of ordinary users.

To spread the word about the benefits of SDNs, IEEE formed a new working group in May to help promote the technology through various activities, including a conference. The Software Defined Network Group is sponsored by the IEEE Future Directions Committee, the institute’s R&D arm.

“An SDN is basically a network of equipment that decouples the hardware from the software,” says Antonio Manzalini, chair of the working group. “It’s a way for operators to create a network out of other components, such as devices, machines, sensors, and ‘smart things,’ which currently aren’t considered network elements.” Manzalini is the project manager at the Future Centre, part of Telecom Italia’s strategy department, in Turin, Italy. An expert in software-defined networks and network-function virtualization, he focuses primarily on the development of new networks and services.

“We want to share information about SDN because a lot of confusion surrounds the technology,” he says. “Our goal is to reach a common understanding on what SDN means and its potential impact on future networks.”

A PRIMER

Part of that confusion is that the terms software-defined networks and software-defined networking are being used interchangeably. But they are different processes, according to Manzalini. Software-defined networking (a derivation from an intelligent network) uses software to architect existing network resources in a specific way, basically doing the same thing that today is being done in the network itself, with a preprogrammed (and often quite flexible) strategy, he says. Software-defined networks, on the other hand, are networks created out of a variety of elements that usually are not even considered as such.

Another important aspect is the virtualization of the network functions (for example, L4 to L7 switches used for load balancing among groups of servers) “This is complementary to SDN and does not depend on it,” he says. “The two concepts should not be mixed, even though they can be combined in several ways that can potentially create a great value.”

Computer networks of yesteryear evolved from an Internet architecture based initially on the well-known end-to-end principle: Packets were forwarded through the network, from source to destination, mainly unmodified.

In today’s networks, traffic crosses several so-called middleboxes—computer network devices that transform, inspect, filter, or otherwise manipulate traffic for purposes other than packet forwarding. Examples include wide-area network optimizers, network address translators, performance-enhancing proxies, application-specific gateways, intrusion detection and prevention systems, and all sorts of firewalls.

These specialized network functions are based on hardware built for purposes that are typically closed and expensive to maintain. Not only are these middleboxes contributing to the so-called “network ossification,” meaning that installations are difficult to modify, but they also cost a lot to maintain, Manzalini points out. Removing or even reducing the number of middleboxes with SDNs and virtualization would have several advantages, not the least of which are cost savings, service improvements, and increased flexibility.

Another factor fueling interest in SDNs is that over the years, large amounts of processing, storage, and communication-networking resources have been moved to what Manzalini calls the “edge” of current networks. In this context, SDN will help companies offer products and services embedded in communication capabilities.

“The edge of current networks would be turned into an area of multiple, interacting networks of virtual resources,” says Manzalini. “Imagine having thousands of sensors not yet connected to the Internet that could detect a variety of conditions. SDN would allow you to manage these as well as a variety of other entities, such as your smartphone, the communication system in your car, or whatever else is around you. You can instruct the network of sensors to detect and deliver data to one area where you could create a new application. For businesses, SDN provides the possibility of creating software networks for a variety of business needs.

“SDN would provide enormous flexibility, just like the OS has provided to the computer industry,” he continues. “Borders between networks and data centers would disappear. The physical location of data centers would no longer be relevant, creating a completely virtual environment.”

LET’S MEET

To better understand the technology opportunities, as well as the context in which these may play out, the working group has organized the Software-Defined Networks for Future Networks and Services conference, to be held from 11 to 13 November in Trento, Italy. The deadline for article submissions is very soon, 13 August.

“As an outcome of the conference, we hope to produce a sort of manifesto for SDN that has IEEE’s stamp of approval,” Manzalini says. It could lead to a website or a journal.

The group is also working on understanding the status of standard hardware and software and the advances needed to implement and exploit SDNs. “The edge of the network is where intelligence has already started migrating, and it is where innovation is more urgently needed to overcome the network ossification,” he says. “The incredible amount of processing storage and communication resources moving to the edge is creating the favorable conditions to exploit SDN innovation at its best.”

In addition, regulatory issues must also be addressed. “We want to know the regulatory aspects of the new business models and the economic sustainability issues for new networks and services,” Manzalini says.

Learn More