"I want to liberate and simplify the world of networks."
It is said that SDN will bring the kind of innovation to the world of networking only seen once every 20 years.
NEC was the first in the world to release products supporting the OpenFlow technology that enables SDN. Atsushi Iwata, also known as Mr. OpenFlow, is involved in both OpenFlow technology research and product development. Here he discusses his thoughts on new value creation, and the things he is particular about as a researcher.
SDN is a buzzword that will redefine networking.
--First, please explain SDN (Software-Defined Networking), which is attracting attention as a concept that will redefine networking.
Iwata: Put simply, SDN is the control of networks using software. Until now, you had to individually configure the many network devices such as routers that exist in large numbers on a network. This meant that only those with specialist knowledge or know-how could operate and configure networks. SDN is a technology for controlling entire networks by introducing a software layer above network devices. This makes it possible for even those without specialist knowledge or know-how to control and operate networks. One simple analogy is that this provides the world of networking with the same kind of convenience as a car navigation system, which avoids road with heavy traffic and automatically selects and informs you of the best route even when you don't know the way.
Another feature of SDN is the optimization of networks to accelerate application performance. A variety of applications are run at workplaces such as offices, including database servers and web servers. In conventional networks these functions are merely connected, but SDN makes it possible to increase processing speed by a factor of 10 by optimizing the network to extract maximum performance from each application. In short, SDN is currently gaining prominence as a technology that creates significant value for both those who operate networks, and end users who utilize ICT systems.
--Next, tell us about OpenFlow, a key technology for implementing SDN.
Iwata: It is often said that OpenFlow is a technology for the centralized control of networks using software as a single brain. However, this is merely one aspect of OpenFlow. Using software it is possible to control networks freely and flexibly without being tied down by physical limitations. It is also easy to program open programmable networks that are free from the constraints of the particular specifications of each network device vendor. I believe the essence of OpenFlow lies in these two points. For example, OpenFlow technology enables the optimal control of a network by company or building, as well as by business unit or individual.
--What are the reasons for the shift to open programmable networks like SDN?
Iwata: I think there are two main reasons. The first, is the business-related network issues of customers. Until now, when adding business categories or providing new services at data center providers and companies, network updates and configuration required a great deal of time and money. Issues directly linked to operation also occurred, such as not being able to launch swiftly when providing a new service to the market. Even though resources such as servers can be readied quickly due to progress in virtualization, changes to networks take several weeks. The fact there was a growing need for a solution to these issues was one reason for this shift. The second reason was the development of bulk data application, as typified by cloud computing and big data. There was a need for innovative networks able to process enormous quantities of data at high speed and provide feedback in real time. NEC has a proven track record of providing solutions in a range of industry sectors, and from this experience we were quick to identify customers' network issues and needs. NEC is also involved in the construction and operation of data centers, and had realized the necessity of new network technology able to support things like the application of cloud computing. Staying ahead of these trends of the times, NEC was among the first to initiate research into new network development.
Working towards innovation in networking only seen once every twenty years.
--Tell us what triggered the research and development of OpenFlow technology.
Iwata: When I visited Stanford University in the United States in 2006, Professor Nick McKeown of the computer science department approached me with a proposal. It involved research activities aimed at developing new network technology from scratch (the Clean Slate Program), rather than revamping existing network technology. At this time research called new-generation networks was being discussed in Japan, and having begun research into new networks already, NEC agreed to the proposal from Stanford University and began joint research. This joint research covered many topics. Back then the term OpenFlow still didn't exist, but research and development into what served as the origin of OpenFlow was one of these research topics. The joint research on OpenFlow at the time contained seeds that would help realize SDN, so we engaged in frequent discussions based on these, while gradually establishing the system.
--What specific forms did the joint research into OpenFlow between NEC and Stanford University take?
Iwata: Basically, we embedded OpenFlow-enabled software into a conventional switch, programmed a new network control application into the controller, and used actual networks to perform trials and work on the completeness of the specification. More specifically, we first modified just the software functions of an existing switch manufactured and sold by NEC to support OpenFlow in a research lab in Japan. Together with NEC's business division, we created a prototype that would serve as the original OpenFlow-enabled switch. Then, we sent the prototype built in Japan to Stanford University in the United States, and performed repeated proof-of-concept tests using networks. At first we performed tests with a network we used and were familiar with, then we gradually expanded the proof-of-concept tests to a variety of buildings on campus, and after that a number of other universities, building up our test results.
A range of data gained through the proof-of-concept tests was sent back to Japan, and used to improve the software and upgrade the prototype. This upgraded prototype was then sent to the United States, where the proof-of-concept test cycle would begin again. This demonstrates how much work was involved. Running an actual network to gradually improve the switch functions and performance could be compared to the development of Formula One cars, which have their performance tuned up based on actual race data.
The world's first to successfully create a product enabling SDN.
--Tell us about development of the "UNIVERGE PF Series" which NEC successfully brought to market as the first OpenFlow-enabled product in the world.
Iwata: In 2009, after about two years of joint research with Stanford University, the road to commercialization was in sight, so we asked the business division to consider the next step. Of course, at the time we had no idea how an OpenFlow-enabled product would be received by the world. NEC had also been involved in the manufacture and sale of network devices up to that point, as well as a range of network construction and operation as a system integrator. We wondered about the kind of impact that new OpenFlow-enabled products would have on these existing areas of our business. After looking at factors such as the new value we could provide to customers, where NEC's business would be in 10 years' time, and growth of the global market, we arrived at the conclusion that we should take the plunge into commercialization, and throw our weight behind the area of OpenFlow. Then, after two years of product development, in 2011 we at last succeeded in developing the world's first OpenFlow-enabled product, the "UNIVERGE PF Series."
--Can you share any inside stories or tales of struggles you faced during development of the "UNIVERGE PF Series"?
Iwata: For the project set up for product development of the UNIVERGE PF Series, development members from a range of departments came together and formed a team. Basically, this was a mixed team of seasoned professionals in the areas of IT and networking. To start with, IT pros and networking pros use different terminology. The culture and specifications for conducting tests are also different. So the first thing we did was sort out the lingo we used. There were all sorts of raucous exchanges in the beginning, with clashes of ideas and opinions. Meanwhile, we gradually recognized each other's strengths, and applied the best resources to each area. As a result of this fusion of the strengths of IT and networking, we created a world-leading product.
--In 2011, around the time you successfully commercialized OpenFlow, what was the state of other initiatives around the world to realize SDN?
Iwata: In 2011, when NEC released the world's first OpenFlow-enabled product, the "UNIVERGE PF Series," a push to standardize OpenFlow technology began around the world. The Open Networking Foundation (ONF) standards organization was founded, with members consisting of major telecommunications carriers, providers, and ICT vendors from around the world. These included Japanese companies such as NTT and NEC, as well as overseas companies including Yahoo!, Google, Facebook, Microsoft, as well as Deutsche Telekom. Efforts towards the standardization of OpenFlow have now spread around the globe. You might say that NEC led its generation in a number of ways, including OpenFlow research and development, commercialization, and standardization.