TTU researchers lead next revolution in online data sharingIn the same spirit that universities showed in launching the Internet, Tennessee Tech University is leading the revolution in how scientists may soon share massive amounts of data over different computer networks.
In support of this vision, the U.S. Department of Energy has committed more than $1.2 million in funding, of which $525,000 directly supports Nasir Ghani's research at TTU. This will allow scientists to send huge data files — ones that would normally take days and weeks to transmit and reach their destination — in hours or less.
Ghani, an associate professor of electrical and computer engineering, said that there is a pressing need in the scientific community to advance the capabilities of today’s networks.
"Think about the limitations of your current computer e-mail systems to attach anything larger than 10 megabytes and send it to someone," said Ghani. "Now imagine the limitations scientists face when trying to send files more than a million times larger than that."
Ghani explained that most individuals are used to considering a gigabyte as the largest unit of memory used on their computers or other electronic devices. However, today many mid-sized companies and universities commonly operate with data on the scale of terabytes, equivalent to 1,000 gigabytes. The DOE, working at a level of large-scale scientific computing now commonly referred to as "e-science," is further seeking to operate efficiently on the scale of petabytes and beyond, i.e., millions of gigabytes.
"So if you are a scientist in Chicago collaborating with someone at Berkley on a project, you want to be able to share data in hours, not days or weeks," he explained. "And frankly, today it is still cheaper and faster to place very large amounts of information on some storage medium such as a DVD and fly it across the Atlantic than it is to transmit it through different computer networks."
Ghani described emerging DOE systems as "networks on steroids." They include DOE's Energy Sciences Network, a live nationwide network that delivers mission-critical services, and UltraScience Network, a smaller experimental optical network. The National Science Foundation is also funding various networking testbeds, including the Hybrid Optical and Packet Infrastructure (HOPI) project and the Dynamic Resource Allocation via GMPLS Optical Networks (DRAGON) project. Finally, the Internet2 organization operates a large high-speed backbone that interconnects the nation’s universities.
However, he says these individual networks alone cannot provide the massive inter-connectivity needed for e-science researchers working in diverse areas such as high-energy physics, astronomy, climate change, nanotechnology and the like. Hence the DOE wants to leverage its existing technologies as much as possible and is asking Ghani and his team to develop an umbrella architecture to blend them together in order to vastly boost overall data-transfer capabilities.
The challenge for TTU researchers is to develop standards and algorithms that can be used across diverse network types. Ghani and his team at TTU will leverage their advanced state-of-the-art simulation and network performance capabilities to rapidly evaluate new architectures. This work will fund several graduate students.
"We will study the different technologies available, blend those in different ways and simulate how the network will respond," said Ghani. "Then we will choose a few of the best strategies and have our research partners build and demonstrate what we have designed."
Led by TTU, the other co-principal investigators are Tom Lehman from the University of Southern California Sciences Institute East and Rick Summerhill from Internet2. Other collaborating organizations also include the DOE, NSF, and Oak Ridge National Laboratory.
Ghani pointed out that just as university researchers developed the Internet browser to share information, academic institutions, instead of commercial companies, are most likely to continue to lead and pursue this work.
"Currently, few commercial operators are demanding this scale of networking; it's just too specialized and there's not a lot of potential for carriers to make a large profit yet," he said. "Moreover, current residential applications such as high-speed Internet or even HDTV don't require anywhere near this amount of bandwidth capacity.
"However, the pursuit of this technology will inevitably benefit us all in the longer-term as commercial applications we may not yet envision will emerge," he said. "Plus, this project will train a highly-skilled cadre of networking scientists and help maintain America’s preeminence in this important field."