Software Defined Networking – SDN
Several months ago six of the world’s largest tech companies – Google, Microsoft, Facebook, Yahoo, Verizon and Deutsche Telekom – joined forces to form the Open Networking Foundation (ONF), which will advance the development of a new open source networking protocol called OpenFlow. What exactly is OpenFlow, and why would these huge companies throw their collective weight behind it?
The ability of a network operator to create custom functions applicable to his own network, and then apply those functions to switches from multiple vendors, is the true
promise of SDN. OpenFlow allows a customer to programmatically control their network, over an industry-standard interface, using the same distributed system libraries and packages they use to orchestrate the rest of their infrastructure. The two key points here are ‘customer programmatic control’ and ‘industry standard.
OpenFlow enables networks to evolve, by giving a remote controller the power to modify the behavior of network devices, through a well-defined “forwarding instruction set”. The growing OpenFlow ecosystem now includes routers, switches, virtual switches, and access points from a range of vendors.
In today’s packet networks, a router/switch is both the control element which makes control decisions on traffic routing, as well as the forwarding element responsible for traffic forwarding, and both these functionalities are tightly linked (Fig. 1a). Housing control and data functions in the same box makes routers complex and fragile, quite unlike the streamlined routers envisaged by the Internet pioneers. Today, a backbone router runs millions of lines of source code, and a plethora of features in software and hardware.
Transport networks are similar. While traditionally they have had a separation between a circuit switched data plane and a packet switched control plane, this control could reside within the box (Fig. 1b) or outside the box with proprietary interfaces (Fig. 1c). Additionally, out-of-box-control may not even be a distributed control plane, but more likely an Element Management System (EMS) / Network Management System (NMS) hierarchy. Those that desire the former are headed towards the same problems seen in packet switched networks today.
OpenFlow advocates a clean separation between the data plane and the control plane in packet or circuit networks (Fig. 2). Because the data plane is typically implemented in hardware, OpenFlow provides the control plane with a common hardware abstraction. A network (for example an autonomous system) is managed by a network-wide operating system (e.g. NOX]) running on multiple software controllers (Fig. 3), that controls the data plane using the OpenFlow protocol.
NOX is an open platform for developing management functions for enterprise and home networks. NOX runs on commodity hardware and provides a software environment on top of which programs can control large networks at Gigabit speeds. More practically, NOX enables the following:
- NOX provides sophisticated network functionality (management, visibility, monitoring, access controls, etc.) on extremely cheap switches.
- Developers can add their own control software and, unlike standard *nix based router environments, NOX provides an interface for managing off the shelf hardware switches at line speeds.
- NOX provides a central programming model for an entire network – one program can control the forwarding decisions on all switches on the network. This makes program development much easier than in the standard distributed fashion.
This video from last year’s Structure Conference, Nick McKeown of Stanford University explains the concepts behind openFlow and the ways it might change the way networks are built and customized.
Open Networking Summit set to explore software-defined networking
For three days this October, a group of computer networking industry heavyweights and academic researchers will assemble at the Li Ka Shing Center at Stanford University for the Open Networking Summit, the first public industry event exclusively focused on a new paradigm known as software-defined networking (SDN) and OpenFlow.
The Open Networking Summit offers a day of hands-on tutorials plus two days of keynote and panel sessions featuring networking thought leaders and influential media.
The names expected to be in attendance reads like a who’s-who of the computer industry: Google, Facebook, Verizon, Yahoo!, Microsoft, Internet2, Nicira, Big Switch, Cisco, HP, Juniper, NEC, Ericsson, Arista, UC Berkeley, Princeton, Stanford, Georgia Tech, and Indiana University. Also in attendance will be network and cloud providers, equipment vendors, chip vendors, virtualization vendors, startups, venture capitalists, and investment banks.
Most industry participants are members of the Open Networking Foundation, a non-profit industry consortium whose mission is to standardize OpenFlow and advance software-defined networking.
“The Open Networking Summit will — for the first time — bring together the people who want to make SDN and OpenFlow succeed, revolutionizing the field of networking,” said Urs Hölzle, a Senior VP of Engineering at Google who is also Chairman of Open Networking Foundation board. “OpenFlow and SDN will accelerate innovation of the Internet infrastructure — just as the introduction of the PC revolutionized the computer industry in the 1980s.”
The Summit itself is the culmination of nearly a decade of close partnership between industry and academia to rethink networking.
“OpenFlow/SDN is a great example of technology transfer from a university to industry,” said Nick McKeown, a professor of computer science and electrical engineering and faculty director of Clean Slate Program at Stanford.
Starting in 2003, the National Science Foundation funded a series of programs to rethink the Internet architecture. One such program (known as the “100×100 Project for Clean Slate Design”) funded the research of Stanford PhD student Martin Casado. Working with professors at Stanford and Berkeley, Casado started to think how networks could be redesigned, as if from a clean slate, to be more secure, more dependable, and easier to manage. The key ideas from his work led to Software Defined Networking and OpenFlow — a new way to control network switches and routers.
With further funding from the National Science Foundation, the concepts blossomed into the networking substrate of the Global Environment for Network Innovation (GENI), a nationwide proof-of-concept project that now links nine U.S. universities, including Stanford, University of Washington, Indiana University, University of Wisconsin, Georgia Tech, Clemson, Rutgers, and Princeton.
GENI allows researchers to test new ideas at scale with real traffic. Software Defined Networking and OpenFlow became an important part of the GENI backbone, helping GENI to achieve its goal of experimentation at scale with real applications and users.
Soon thereafter, the networking community began to take note. A group of information technology giants — including Google, Facebook, Microsoft, Verizon, Deutsche Telekom, Yahoo!, HP, NEC, Dell, Juniper, among many others — embraced OpenFlow.
The Open Networking Summit will be a major milestone in broader acceptance of OpenFlow/SDN.
“NSF has a long history of investing in foundational computing research that has led to the transfer of knowledge from lab to practice,” said Farnam Jahanian, NSF Assistant Director for Computer and Information Science and Engineering. “We are thrilled to see that software defined networking and OpenFlow have became an important part of the GENI backbone, and we look forward to future innovations that build on these important investments.”
“The Open Networking Summit is an important part of the technology transfer of OpenFlow/SDN. Here we have a university research idea funded by government now becoming an important new direction for industry,” said McKeown. “Much of the credit for where we stand today rests with the National Science Foundation and its early intellectual backing and funding of these ideas.”