ENGINEERING

To the uninitiated, dark fiber sounds a bit mysterious, maybe even daunting. But to network researchers dark fiber means an environment removed from the demands of a production network; a place in cyberspace where they can conduct experiments and explore the future of networking removed from the demands of a commercial network and partitioned off so that one experiment won't interfere with another.

Meet BEN, the Breakable Experimental Network, a dark fiber, experimental network test bed where researchers can push the limits of networking technology unencumbered by the day-to-day requirements of a production network. Created and deployed by RENCI's network research division in partnership with Infinera, an optical network hardware provider, BEN exists apart from the heavy traffic and bottlenecks of research networks and the commercial Internet. It's "breakable" because it doesn't demand the 24-hour reliability of a research or commercial network and can therefore be taken out of commission if required by an experiment.

The BEN research team aims to create a sliceable, highly programmable, network provisioning system that can dynamically and reliably deliver the resources needed for advanced networking research.

Ilia Baldine, RENCI's manager of network research projects, describes BEN as a "sandbox" that gives researchers a controlled environment in which to conduct research. Networking experiments performed on BEN will be repeatable, he says, and therefore more meaningful to the research community.

"A modern-day network consists of multiple layers that operate in concert," says Baldine. "In order to assure quality of service over multiple layers, you need to understand all the layers of networking down to the fiber level. What we have with BEN is a framework where different experiments can be conducted over different layers without messing things up."

Of ORCA and GENI

For BEN to work, it needs common software to manage its distributed network resources. Duke University's Open Resource Control Architecture (ORCA), developed by Jeff Chase, a professor in the Duke computer science department, serves as BEN's resource control system. ORCA was designed to securely and efficiently manage heterogeneous distributed resources and to simplify the process of resource allocation in a distributed system. ORCA "leases" resources to users without imposing any particular structure on them, making the resources more accessible to researchers and giving them greater flexibility.

As Baldine explains, ORCA is a control framework that makes advanced networking research possible. ORCA, combined with the BEN experimental test bed, is precisely the kind of infrastructure that the Global Environment for Network Innovations (GENI) supports. GENI is a project sponsored by the National Science Foundation (NSF).

"GENI's current phase is about finding the best features of a control framework for scheduling and controlling resources and measuring capabilities for large-scale experiments," says Baldine. "This is what we were already planning to do with BEN and ORCA so it makes sense for us to be part of GENI."

Through the GENI project, RENCI and Duke are deploying the ORCA control framework on BEN as a GENI "cluster," a prototype version of what may one day be a national-scale network research infrastructure for large-scale experiments on issues such as distributed systems, storage systems and how to manage large-scale distributed resources to avoid performance bottlenecks. ORCA/BEN is one of five integrated clusters being developed as GENI prototypes by teams across the U.S.

Intuitive troubleshooting with MIN

Add visualization to a research project and you generally have a "wow factor" as well as an important research tool. BEN is no different and its wow factor boasts its own acronym -- MIN, for Multi-touch Interactive Network.

MIN does what all successful visual analytics projects do -- it makes data easier to understand by converting columns of numbers, charts and tables into intuitive, color-coded images. Chris Heermann, a senior research scientist at RENCI and Xunlei Wu, a senior visualization researcher at RENCI's Duke engagement center, created a visualization of BEN, which uses real-time performance data from devices deployed in BEN. Heermann and Wu went a step beyond visualization, developing MIN as an interactive tool that runs on the large multi-touch display wall at the Duke-RENCI center.

"If you use the analogy of your iPhone, MIN is a 13.5-foot by 5-foot iPhone showing all the functionality of a dark fiber network on its screen," says Heermann. "Typically, tracking network problems is difficult and involves a lot of charts and tables. With MIN, you can walk up to the visualization and touch it, detect problems through things like color coding, and magnify your view so you can see what's happening not just at the POP level, but down to the device level."

Heermann sees MIN as both a practical tool for monitoring network traffic and spotting performance problems and as a research project that could change how network engineers and systems administrators interact with networks.

"It's a new and very sophisticated way of monitoring a network that lets you put your hands on the network and all its devices," he says.