Skip to content

ALPHA PROJECTS | Contents | Next

By Sid Karin
NPACI and SDSC Director

Beyond Isolated Central Systems:
Integrated Grid-Based Computing

A s we enter the 21st century, the traditional model of stand-alone, centralized computer systems is rapidly evolving toward grid-based computing across a ubiquitous, continuous, and pervasive national computational infrastructure. In the NPACI vision, researchers collect data from digital libraries and experiments, analyze the data with models run across the grid, visualize and explore those data over the Web, and publish the results to the scientific community in digital libraries. While other definitions of the grid may emphasize different factors, in all of them grid computing is far more than just an easier way to access the same old resources--it involves an ongoing process of integration in which the whole is far greater than the sum of the parts as researchers attack new problems in ways that are fundamentally different from past methods.

I should make clear that I'm not at all saying that the need is less for large-scale computing resources. Quite the contrary, the need is greater than ever, as requests for allocations on NPACI resources consistently run more than twice the available time, a frustrating impediment for many researchers and strong evidence that such resources could usefully be a higher national priority.

NPACI and SDSC continue to build on established capabilities to provide the maximum possible computing power to the academic community. Indeed, NPACI's Blue Horizon--an IBM RS/6000 SP at SDSC that is the nation's most powerful academic computing resource--has recently been upgraded from 1 teraflops to 1.7 teraflops capability.

Grid computing
involves an ongoing
process of integration
in which the whole
is far greater than the
sum of the parts.

The point, though, is that high-performance scientific computing resources can no longer be seen as isolated systems that exist like islands. Rather, they must be viewed in the larger community context of grid-based computing. But the promise of grids--enabling more difficult scientific problems to be attacked in larger collaborations--comes with a challenge: greater complexity, transcending any one field or the expertise of any one researcher, and often any one department or even institution. Modern research is now irrevocably multi-disciplinary and multi-institutional as researchers cooperate in new ways to take advantage of these unprecedented opportunities.

Perhaps nowhere within NPACI are the opportunity and challenge of grid computing more apparent than in the alpha projects. Each of NPACI's seven alpha projects, which are highlighted in this issue of enVision, involves a core element that focuses on a unique aspect of grid computing: The Bioinformatics Infrastructure alpha project focuses on analyses that are data-intensive; Protein Folding on large-scale computing analyses; Telescience on integrating instrumentation with data and computation; Multi-Component Models on linking large computations involving a wide range of scales; and Scalable Visualization Toolkits on visualization of large data sets with multiple scales. Added to these are two new alpha projects: Adaptive Computations, which focuses on advanced computational tools for biological flow simulations, and Cellular Microphysiology, which is extending neuroscience research from the single neuron to the larger neural environment through efficient distributed technologies.

Of course, beyond these alpha projects are numerous other projects in grid computing that involve NPACI and SDSC researchers. Of particular note are two recently funded projects in NSF's major Information Technology Research (ITR) initiative. "Virtual Instruments: Scalable Software Instruments for the Grid" will develop grid-based software to steer simulations of cellular function. Led by Fran Berman of UC San Diego, the project also includes NPACI partners at the Salk Institute and the University of Tennessee. A second project, "GriPhyN: the Grid Physics Network," will lay the groundwork for a computer data grid of unprecedented speed and power, supporting four large physics experiments. Co-leaders of this project, the largest ITR award, are Paul Avery of the University of Florida and Ian Foster of the University of Chicago, with participating NPACI researchers at Caltech, UC San Diego, the University of Southern California, and the University of Pennsylvania.

These groundbreaking efforts, as well as a great many others within both NPACI and our sister partnership the Alliance, are constructing the fabric of grid-based computing that is connecting high-performance computing, data, and visualization resources through high-speed networks into broadly collaborative scientific applications. And while high-performance computing resources are in as much demand as ever, no longer are they isolated centers. They have become transparent parts of grid-based applications that are accelerating the pace of science and opening fundamentally new avenues of discovery. *