When long-haul truckers hit the road with their freight, they often rely on GPS to help them find the best possible routes and adjust in real time. So why can’t the online systems for transferring large datasets, which typically rely on predetermined settings and don’t adjust in real time, take a similar approach?

That’s what Dr. Md Arifuzzaman, an assistant professor of computer science at Missouri S&T, aims to solve. Backed by a $175,000 grant from the National Science Foundation, he is working to develop a new data transfer tool.

“My goal is to build a system that moves data more efficiently and makes better use of the network resources connecting high-performance computing centers by learning from past experience and adapting on the fly,” he says.

The system Arifuzzaman is developing will include offline training of artificial intelligence programs that learn by practicing with real network data in simulated network conditions and allowing it to learn and make decisions throughout its journey as well. This work is a continuation of his doctoral research on the same topic, which earned him top honors in the international Data Mover Challenge competition.
 
“Most systems either don’t adjust well or take a long time to learn,” he says. “By training general strategies in a simulator first, the hope is this system will be elastic and can flex and adapt quickly to new network environments. It is all about faster convergence of learning agents and maximizing unused network capacity in a way that does not negatively impact other system processes.”

Arifuzzaman says big data transfers may not often be on the public’s radar, but it is essential for their everyday lives in ways they may not realize.

“Even though it works behind the scenes, this technology will touch everyday life by doing things like speeding up online services that could support breakthroughs in medicine and science,” he says.

Looking ahead, Arifuzzaman says he intends to share the technology as an open-source tool so other researchers and institutions can make changes and improvements for their own data transfer needs.

“Personally, I have a big vision for where this can go,” he says. “I want to build the next-generation data transfer tool — something that’s robust, adaptive and widely adopted by the scientific community. This project feels like a strong step in that direction, and that’s what makes it so exciting for me.”

About Missouri S&T

Missouri University of Science and Technology (Missouri S&T) is a STEM-focused research university of over 7,000 students located in Rolla, Missouri. Part of the four-campus University of Missouri System, Missouri S&T offers over 100 degrees in 40 areas of study and is among the nation’s top public universities for salary impact, according to the Wall Street Journal. For more information about Missouri S&T, visit www.mst.edu.