RENCI’s Breakable Experimental Network means new research opportunities at UNC
CHAPEL HILL–A laboratory can be a physical space—such as the Networking Laboratory in the computer science department at UNC Chapel Hill. Or it can exist in cyberspace, like RENCI’s Breakable Experimental Network (BEN), a regional optical network test bed for experiments with disruptive networking technologies.
Combine BEN with the expert faculty and resourceful students in the aforementioned Networking Lab and you’ve created a recipe for cutting-edge research that could transform the Internet and its ability to transport ever-increasing loads of data.
“We have this lab and we have a 10 gigabit per second network, but it is a really small network. A real research network would involve scientists sharing data over much greater distances,” says Jasleen Kaur, an associate professor in UNC’s computer science department who studies packet-scale congestion control on high speed networks with the goal of implementing a new transmission control protocol (TCP) that would allow data to travel at speeds of 1 terabit (1 trillion bits) per second or faster.
Kaur’s research team has developed a system called RAPID, a paradigm shift in how networks probe for the spare bandwidth that allows large datasets to be transported quickly without impacting other network traffic. The paradigm relies on two principles: fine-scale probing that creates finely controlled packet sending times and estimates the available bandwidth based on the delays experienced; and probing-without-overloading, which finds a wide range of bandwidth rates within a single round trip time (RTT—the time it takes to send a signal and receive a response, sometimes called a ping), which prevents the network from becoming overloaded.
BEN enhances the research team’s work in two ways, according to Kaur.
“By connecting to BEN we have a 10 gigagbit path that goes from our department lab to the BEN PoP at RENCI, it goes to Duke, it goes to NC State and all the way back and we can create many interesting topologies,” she says. “The BEN PoP at RENCI can also be connected to the NLR (National Lambda Rail, the high-speed research network connecting universities across the U.S.). We’ve been able to set up a path that starts at our lab, goes through BEN to the NLR through several cities across the U.S. and comes back to BEN and back to our lab. Now, using our own machines as clients and servers, we can emulate a really long distance network path.”
The science of networking
Research Professor Don Smith came to UNC Chapel Hill from IBM in 1997 with the goal of “bringing science to networking.” New protocol ideas, routers, switches, etc., must be evaluated, he explains, but in many cases the evaluation process doesn’t live up to the requirements of sound scientific methodology. Too often network testing environments fail to create realistic conditions that reflect what happens on the Internet, says Smith. In addition, network experiments are often done in uncontrolled environments, which makes it difficult for other scientists to reproduce experimental results.
Smith and Kevin Jeffay, Gillian Cell Distinguished Professor in computer science at UNC Chapel Hill, have spent years developing a networking research environment that allows them to reproduce the traffic conditions of a large production network in the controlled conditions of the networking lab. The researchers’ methodology involves using packet traces from live networks to create synthetic models of network workloads. They can then run their modeled workloads in the lab to test new protocols in a realistic, but controlled, situation.
The methods used to reproduce network conditions scale up to the levels of the most advanced multi-gigabit research networks, such as the Internet2 backbone. With the addition of BEN, the traffic generating system can be run on a much larger controlled network infrastructure (the BEN test bed) and an even larger working research network (NLR).
“That’s very important for evaluating something like RAPID because we need to know how RAPID responds to dynamic changes in the traffic,” says Smith.
It also means more visibility for UNC’s traffic generation methodology and a trend toward more researchers reproducing that methodology in their own work.
“I think it’s going to make us one of the premier experimental networking facilities in the country,” says Smith.
For more information:
UNC Chapel Hill Computer Science Department
RENCI supports the Innovate@Carolina Roadmap, UNC’s plan to help Carolina become a world leader in launching university-born ideas for the good of society. To learn more about the roadmap, visit http://innovate.unc.edu.
