Software Defined Networking: The next big thing?
In the new
edition of Embedded Systems Design Magazine, Daniel Proch,
in "Controlling network flow with the OpenFlow
protocol", describes a new network protocol developed to solve a
potential Internet traffic jam of monumental proportions. The protocol was
developed by Internet backbone equipment providers and major technology
companies to head off a "perfect storm" of events they felt could bring
everything to a sudden stop if the very architecture of the public network was
First there was the continuing improvement in the Internet backbone speeds, with data rates in 10 to 100s of gigabits per second becoming common. Second, gigantic server farms now make the idea of "cloud computing" – the storing and processing of huge amounts of data remotely rather than locally – a reality. Third, there was the final transition to IPv6 and the enormous number of URLs now available – enough for not only every PC, but for every mobile phone for every human on earth. And with 6LoWPAN, this connectivity extended into a variety of wireless sensor applications as well.
To head off these problems, companies such as Cisco, Deutsche Telekom, Facebook, Google, Microsoft, Verizon, and Yahoo, to name a few, came together a year ago and formed the Open Networking Foundation. Their aim: creation of a virtualized "software defined networking" architecture flexible enough to manage such problems and low cost enough to be implemented quickly and broadly.
Fortunately they did not have to wait years for such a scheme to emerge. One that was already at hand, the OpenFlow protocol, a multi-university cross-platform scheme, extends the concept of network virtualization used in many closed networks to the broader Internet. Network virtualization is not new and the approaches have been described in many articles on Embedded.com and in ESD, including:
Using virtualization to consolidate data traffic on a single network appliance
Network I/O virtualization and the need for network I/O processors
Using PCI Express I/O virtualization to pool network resources
Even before a commercial version of the protocol is formalized, a number of products have become available from companies such as Big Switch, Broadcom, Cisco, Hewlett Packard, IBM, Marvell, NEC and Netronome - many of them hybrid designs to allow smooth flow from the old to the new.
Two white papers I think you will find useful are "Designing Datacenter Networks: Alternative Approaches with OpenFlow", and "IBM White Paper: OpenFlow - Next Generation in Network Interoperability". And here's a list of technical papers published over the past year on using OpenFlow in a variety of networking environments. If you want to get down to the specifics of this protocol and some of the alternatives being discussed, plan on attending the second annual Open Networking Summit in Santa Clara, Ca., April 16-18.
Despite the increasing momentum toward OpenFlow, a lot of skeptical tire-kicking is still going on. According to Proch, while some factions in the industry believe OpenFlow is the next big thing in computer networking, others think it is just the newest fad and will fade away while existing networking technologies and methods continue to be prevalent.
What do you think? I'd like to hear from you with design articles and blogs on OpenFlow: its strengths, its weaknesses, how it can be improved, how you are implementing it in your designs, and if you are not doing so, why not? What alternatives are you investigating?
This article provided courtesy of Embedded.com and Embedded Systems Design Magazine. Sign up for subscriptions and newsletters. Copyright © 2011 UBM--All rights reserved.