In a few unused back alleys of the Internet, researchers are testing radical new ways of transferring information, often at speeds almost unimaginable to the home Internet user.
Internet2, a consortium devoted to developing advanced networking applications and technologies, and the Energy Sciences Network (ESNet), which provides powerful data connections for scientists at national laboratories, universities, and research institutes, are putting together experimental networks on top of dormant networking resources known as “dark fiber.” While the researchers say it will be years before the advances reach individuals and businesses, they think the work will ultimately ensure that the Internet functions smoothly in the future. For example, the experimental networks could allow researchers to update protocols, anticipate security needs, try out better hardware, and look at ways of making networks more energy-efficient.
The organizations worked together on two prototype networks. One transfers data at a mind-boggling 100 gigabits per second. (Google made waves last year by announcing its plan to build a one-gigabit-per-second network for a chosen community.) The second is intended for riskier experiments into network architecture. Researchers want to find protocols that transfer data faster and more reliably, try out innovative network hardware, and experiment with ways to handle difficult security scenarios. In other words, they want to do the type of work that could cause problems for regular traffic if it were done on any sort of shared network.
“When you want to do something disruptive, when you want to try something really radical, you can’t do that on a network that people are trying to actually use,” says Robert Vietzke, Internet2’s director of network services. At the same time, it is useful to test these ideas on real network infrastructure. In the past, he says, researchers would buy spools of fiber, install them in their labs, and try to emulate a national network. Using dark fiber lets them test ideas at larger scales, and bring in real traffic (without disrupting it). This makes it much easier to go from experiment to prototype to actual deployment.
Dark fiber refers to fiber-optic cables that are currently lying unused. “With the dot-com bust, this fiber became available at fire-sale prices,” says Steve Cotter, department head for ESNet. Though the price of dark fiber had risen again in recent years, Cotter says, the economic downturn has made it cheap again. ESNet and Internet2 took the opportunity to lease fiber for the next 20 years.
For the 100-gigabit-per-second network, ESNet and Internet2 added equipment and software to allow the network to function. The experimental test bed, however, which is a separate network, gets left dark—open to whatever equipment or protocols researchers want to bring to it.
As an example, Cotter says, it’s well known that the current protocols for transferring information over the Internet, such as TCP/IP, are creaking with age. The decades-old protocols break down in very-high-bandwidth situations in particular. Cotter expects researchers to test new alternatives.
The dark-fiber test bed provides a way to test new hardware, such as devices like optical routers, which combine optical and electronic components. It could also be useful for security researchers. “It’s very hard to do ‘deep-packet inspection’ at 100 Gbps,” says Cotter, referring to a technique typically used to check for malicious network traffic. He adds that it’s important to anticipate what could happen if spammers or malicious hackers got access to very high bandwidth, and develop ways to defend against that.
Cotter also believes the new networks “open up a whole new area of research” into the power efficiency of network devices. He wants to collect data on real-time power consumption and correlate it with traffic patterns to see when the network itself is becoming an energy guzzler. Data centers have received a lot of attention for the energy they consume, Cotter says, but the energy used to move data around the world is also important.
Lachlan Andrew, an ARC Future Fellow at the Centre for Advanced Internet Architectures at Swinburne University of Technology, says the 100-gigabit-per-second network “will be very useful for people testing end-to-end ‘application layer’ protocols.” That network won’t do much for researchers who want to investigate the deep structure of the network or the hardware used to operate it, but it will be good for those who want to test software in high-bandwidth conditions, Andrew says.
The dark-fiber network, on the other hand, lets researchers investigate networks at any level they want. However, it requires researchers to install their own equipment at many points inside the network to “light up the fiber.” Andrew says that’s a great opportunity for electrical engineers working on networking hardware, but might be an expensive prospect for others.
The new networks “will provide an environment that is realistic, truly state-of-the-art, and flexible for experimentation,” says Steven Low, a professor of computer science and electrical engineering at Caltech who researches network architecture. “Such an environment will be critical for cutting-edge applied networking research.”
Vietzke says construction of both the 100-gigabit-per-second network and the dark-fiber research test bed are well underway. The first stage will be completed by the end of the year. The next stage, which will extend the network to new locations, will be completed within a year.
It’s not clear exactly what will come out of access to the networks, and Vietzke says that’s half the point. “I don’t think you could have imagined that the bandwidth-rich environment [typically available at universities and research centers] could have transformed global politics, commerce, and economics as much as it has in recent years,” he says. He points not only to scientific advances but to dorm-room inventions such as Facebook and Napster. There’s every reason, he says, to expect the next generation of the Internet to be just as disruptive.