Leading the Future of Data Transfer

High in the mountains of Chile, construction has begun on a telescope that can map the entire visible sky in just a few nights. Each panoramic snapshot of the The Large Synoptic Survey Telescope’s (LSST) 3200-megapixel camera will cover an area 40 times the size of the full moon. All of that power comes with a lot of data to be transferred to scientists in the United States at a mind-boggling rate of 100 gigabits per second. That’s where Georgia Tech comes in.

Researchers at Georgia Tech and Florida International University (FIU) have launched the AtlanticWave SDX project to develop a network controller that will prioritize and route large amounts of data on research and education networks without slowing down the internet for everyone else. Sean Donovan, lead software developer and research scientist with IPaT and GT-RNOC, presented the first version of the network controller at a recent conference where he received feedback from network operators and engineers and discussed challenges of the project, including discovering that not everyone understands how bandwidth works.

“You know to get 30, 50 or 100-megabit internet for your home, but you may not actually know what that translates to,” said Donovan. “Thirty megabits is fine if you want to watch Netflix, but if you want to transfer a 50 gigabyte dataset overnight you need to know how long that’s going to take.”

He found that scientists think about data transfers in terms of time, not the amount of bandwidth needed—for instance, do they need the dataset today, or next week? The AtlanticWave SDX team defined new, common protocols to prioritize data flow by time sensitivity, size, or privacy requirements. They’re also working on a more streamlined internetwork connection, which currently takes several network operators one week and a dozen emails to set up. The new system will take seconds.

“That’s the end goal – to make it so that it’s simple for a researcher to use, not just the network operators getting on the phone talking to each other to set up a connection,” explained Donovan.

In the future, AtlanticWave SDX could impact how individual Internet Service Providers (ISPs) prioritize data, with users paying the cost of the bandwidth or extra performance when needed.

"While this current work is focused on scientific use cases, the results will also have important applications to our smart communities projects where we are dealing with increasingly large datasets and complex interactions of different applications," said Russ Clark, co-director of GT-RNOC and a principal investigator for the project.

Georgia Tech and FIU are partnering on the project, in part, because the Institute houses and operates Southern Crossroads (SoX) in Atlanta—a high-speed, large-scale internet network that joins together academic communities in the southeast and around the world. GT-RNOC also develops and tests new network infrastructure. In addition to Clark and Donovan, Georgia Tech PhD student Joaquin Chung, and MS students Ankita Lamba and John Skandalakis are also working on AtlanticWave SDX.

The team is planning an updated demonstration of the network controller during the Supercomputing 2016 Conference in November. The first prototype will be released by the end of the year, with deployments starting in early 2017. AtlanticWave SDX is supported and funded by the National Science Foundation and GENI. The project builds on previous work in Campus SDN and Regional SDX supported by the GENI Project.