University team brings home first place prize for ‘federated cloud’ of supercomputing resources
Until now, using supercomputing to solve research problems or industrial tasks required significant expertise and effort. And if a single supercomputer couldn’t provide enough power, you had to build a bigger one or link two or more together – not a trivial undertaking even for the experts.
The team brought home first place in a contest run by the Institute of Electrical and Electronics Engineers (IEEE), known as the IEEE SCALE 2011 Challenge.
The challenge: to demonstrate scalable computing, that is, the ability to add resources as a task calls for more computing horsepower and to shed them once the task became less daunting.
“We demonstrated how to build ‘federated clouds,’” said Parashar, an effort that involved mixing and matching multiple computing resources, integrating them into a virtual ‘cloud’ that operated as a single large computer, and disbanding resources when they were no longer needed.
“Our goal is to make these federated, high-performance computing clouds more useful to industry,” he said.
Parashar and his National Science Foundation-funded Center for Autonomic Computing (CAC) worked with experts at IBM and the University of Texas to solve a real-world problem: how to extract as much oil as possible from an oil field.
“Oil reservoir optimization is one of the most demanding applications in computing,” said IBM’s Kirk Jordan, a former adjunct professor of computer science at Rutgers. “Modeling the systems requires simulating complex physics and exploring large numbers of parameters.”
The computing competition took place in May at the International Symposium on Cluster, Cloud and Grid Computing sponsored by IEEE and ACM (formerly the Association of Computing Machinery).
The team’s computing needs were provided by IBM “Blue Gene” supercomputers in Yorktown Heights, N.Y., but when more power was needed, the team tapped IBM supercomputers at the King Abdullah University of Science and Technology in Saudi Arabia.
“Accessing such a large supercomputer configuration from an iPad to easily run a complex application makes supercomputing much more accessible to scientists and engineers,” said Moustafa AbdelBaky, a Rutgers graduate student who presented the demonstration at the conference.
Other team members included Hyunjoo Kim of Rutgers; Hani Jamjoom, Vipin Sachdeva, Zon-Yin Shae and James Sexton of IBM; and Mary Wheeler, Gergina Pencheva and Rzea Tavakoli of the University of Texas.
Parashar formed the Center for Autonomic Computing in 2008 to advance the relatively new field of autonomics, which aims to reduce the complexity of building and managing computer systems while making computers more reliable and cost effective. The endeavor seeks to make computers aware of conditions such as excess power consumption, running short on resources like memory and processing power, or sensing attacks from the network, and then equips them to respond automatically with less human intervention.
CAC’s academic partners are the University of Arizona and the University of Florida, and it has 15 industrial partners that include IBM, Intel, Microsoft and Xerox as well as others in the communications, defense and financial industries.
This animated slide deck provides an overview of the demonstration, and an article on the Rutgers School of Engineering web page explains the concept of cloud supercomputing in more detail.
Media Contact: Carl Blesch
732-932-7084, ext. 616
E-mail: cblesch@ur.rutgers.edu