UB's Supercomputing Center Makes Its Mark on Grid2003

BUFFALO, N.Y. -- The University at Buffalo's Center for Computational Research will be a major participant this week in Grid2003, one of the largest public displays of an international computational grid running numerous applications across dozens of sites involving thousands of processors.

For Grid2003, which is held in conjunction with Supercomputing 2003, CCR is acting as a:

* Resource provider, allocating 464 processors -- the largest number of processors of any site on the grid

* An application provider, demonstrating the Shake-and-Bake molecular structure determination algorithm, (developed at UB and Hauptman-Woodward Medical Research Institute), one of only 10 application demonstrations at Grid2003

* Provider to the grid community of CCR's state-of-the-art ACDC Grid Monitoring tool, which provides real-time information on the status of the grid from global and local perspectives.

"Similar to the electrical grid, where users don't know and don't care how electricity is supplied to the outlet that they can plug their microwave oven into, a computational grid is designed to supply a computing platform so that a user doesn't know or care where important scientific and engineering computations are being performed," explained Russ Miller, Ph.D., director of CCR and UB Distinguished Professor in the Department of Computer Science and Engineering, UB School of Engineering and Applied Sciences.

"A computational grid is a collection of heterogeneous, locally controlled resources that are bundled together through software -- called "middleware" -- in order to provide a ubiquitous compute platform for users," he said.

In the future, he added, grids not only will supply computational resources, but also provide access to data, to instruments (such as telescopes) and to people.

The goal of Grid2003 is to integrate, deploy and apply a functional grid across a large number of institutions, including those outside of the U.S.

Miller noted that although UB was not a founding member of the Grid2003 project, CCR was invited to participate due to its strengths in providing critical monitoring capabilities, significant computational power and a world-class application. California Institute of Technology, Argonne National Laboratory, Harvard University, The Johns Hopkins University, the University of Illinois at Urbana-Champaign and the University of California at San Diego are among the other participants.

"It is an honor to be invited to participate in Grid2003," said Miller. "We are on the leading edge of grid computing as a result of producing a working, heterogeneous grid in Western New York. CCR has made significant technical contributions to grid computing that have been recognized by the Grid2003 consortium."

By participating in computational grids such as Grid2003, he explained, and making available its computing resources to the international community, CCR, in turn, gains access to the resources of the other participants. For example, during the past month, CCR has been able to run tens of thousands of computations on this grid.

At Grid2003, UB researchers will demonstrate Shake-and-Bake, the protein-structure determination software package now used throughout the world, which was developed by scientists at UB and Hauptman-Woodward Medical Research Institute (HWI) in Buffalo, and which was listed as one of "The Top Algorithms of the 20th Century."

Miller is principal investigator on a $2 million National Science Foundation grant to produce a grid-enabled version of Shake-and-Bake. This includes developing the requisite "middleware" tools and incorporating data warehousing and data mining for parameter optimization, as well as developing a grid-based interactive 3D version of a tool to display the molecular structures that are determined by a wide variety of methods, including Shake-and-Bake.

"It was important that we deploy a general-purpose grid in Western New York and that we develop critical tools that can be applied in a general fashion to computational and/or data grids," said Miller. Grid2003 serves as a perfect test platform for us to evaluate the progress we have made in terms of grid computing. To date, we have met all of our milestones."

Miller said that one of the greatest challenges in developing grids is the ability to make all of the computational resources, which are at distinct locations and are administered by distinct organizations, behave as though they are a single resource.

"People have to design software that will respect the local computational policies and provide security at a level that meets or exceeds the sites in question," he said. "When a user comes to a portal and fires off a job, the grid must mobilize the resources that will do the job most efficiently and all of that must be done behind the scenes."

Mark L. Green, Ph.D., UB postdoctoral research associate at CCR and research assistant professor in the Department of Civil, Structural and Environmental Engineering, UB School of Engineering and Applied Sciences, is the project's technical leader.

"In a dynamic computing environment, such as a grid, it is important for users and administrators to be able to effectively monitor the environment," he said. "Linking up computational engines, including clusters, supercomputers and special purpose devices, is only the first step in the process. It is critical to be able to know how the system is performing, both locally and globally so that scheduling and turnaround time can be optimized."

According to Green, CCR has provided excellent monitoring tools to the rest of the grid community.

"Computational grids are in their infancy, so it's critical that people are able to view internal resources in their own organizations and those available outside their organizations," he said. "In this manner, one can most effectively analyze the computational load, network bandwidth and data demands."

Concerning Grid2003, Miller says that "working with this talented group of innovators provides significant opportunities for UB to realize scientific advancements in the short-term that will, in the long-term, result in the development of technologies that will be used throughout your home, work and recreational environments," said Miller. "Just as you have a microwave oven in your kitchen as a result of the space program, advances resulting from the computational grid and supercomputing eventually will result in an improved quality of life."