Special Features:

BUILDING BRIDGES
By Kathleen Ricker, NCSA Science Writer

In the summer of 2003, earthquake engineers and computer scientists around the country created an "earthquake" and studied its effect on the structure of a one-story frame building. However, one column of the building was at the University of Illinois; another, at the University of Colorado. Meanwhile, the building's central support was simulated on a computer at NCSA. Amazingly, the structure survived an earthquake 1,000 miles long.

The experiment was called the Multi-site Online Simulation Testbed, or MOST, and combined physical experiments conducted hundreds of miles apart with numerical simulation. It was the first scientific experiment conducted using the NEESgrid, the cyberinfrastructure for NSF's George E. Brown, Jr. Network for Earthquake Engineering Simulation (NEES), which enables researchers at different institutions to remotely interact with each other, collaborate in research, execute experiments, and publish results. Last month, the fully functional NEESgrid was handed over to the NEES Consortium.

"NEESgrid is NSF's first engineering cyberinfrastructure project," said Bill Spencer, who is the Nathan M. and Anne M. Newmark Endowed Chair of Civil Engineering at the University of Illinois and principal investigator for NEESgrid. Spencer said that, by providing common tools that allow individual researchers access to the resources and experiences of a large part of the earthquake engineering community, NEESgrid is making possible research capabilities that have not been available in the past. All of this is accessible through a simple Web interface.

"Rather than having to worry about the software," said Spencer, "NEESgrid allows researchers to focus on the earthquake engineering challenges at hand."

Sridhar Gullapalli of the University of Southern California's Information Sciences Institute agrees. "We are building a community of sites to enable sharing of earthquake engineering expertise and resources," said Gullapalli, who leads the effort to prepare equipment sites to participate fully in NEESgrid. "We are well on the way to enabling cooperative, big science, by getting the sites to work together and by developing a standard way to access NEESgrid services across many heterogeneous systems."

As director of both the Mid-America Earthquake (MAE) Center at UIUC and MUST-SIM, UIUC's NEES equipment site, Amr Elnashai says that NEESgrid will help assure that other researchers will be able to access MUST-SIM more easily and use it to its full capacity. "And as a researcher myself," he said, "even if I am not using our own equipment, I am looking to investigate simulation problems that are extremely difficult to resolve either by integrated analysis or integrated experimentation."

The tools provided by NEESgrid will facilitate research such as Elnashai's, which combines both physical and computational simulation.

How NEESgrid Works

A collaborative NEESgrid experiment typically involves simulating the effects of an earthquake. Data are acquired in a number of different ways, through audio and video feeds, still cameras, and sensors attached to an experimental setup at a NEESgrid node. Through telepresence, data streaming services and a user-friendly browser interface, researchers at different institutions can observe the experiment in progress or, at a later time, work with the data, which is then stored in a central data and metadata repository currently located at NCSA and accessible through the NEESpop, the collection of NEESgrid services installed at each site.

Making it possible to access data in different ways has been a crucial goal of the project. A recently integrated feature of NEESgrid, Data Turbine, provides multiple users with real-time, networked access to the data from an experiment. Charles Severance of the University of Michigan School of Information, data architect for NEESgrid, describes this new capability as a kind of "data" TiVo: "While you're watching the data, it's all being recorded, and yet you can play data backwards and forwards in time, you can skip ahead, you can skip backwards, you can actually pull a whole sequence of data out."

Giving Data A Second Life

In addition to the ability to manipulate data, another feature of the NEESgrid that may prove especially useful is the ability to store data easily in a central repository.

"All my previous experimental projects have one thing in common: that we use maybe five percent of the data we collect," said Elnashai. He noted that if the data his group has collected and worked with is then stored in an easily accessible form, other researchers will also be able to reuse it for their projects or to use it to verify their analytical models.

The NEESgrid data services, the set of technologies and services with which NEESgrid users store and manage data, were the result of a broad and complex collaboration in which NCSA researchers played a large part. The centerpiece of the NEESgrid data services, the data repository, was designed and implemented by Joe Futrelle, leader of NCSA's Digital Library Technologies Group. The repository stores two kinds of data: the raw data generated during an experiment, and metadata, descriptive data which imparts complex, detailed information about the experiment, its design, and the facility in which it is run. The NEESgrid data/metadata team, led by Kincho Law of Stanford University, has developed reference data models for shaking tables and centrifuges.

The Challenges Of Collaboration

NEESgrid has been a breakthrough project in another important way: as a learning experience in conducting a successful large-scale cyberinfrastructure collaboration between two disparate groups of researchers. "NEESgrid has been one of the first serious attempts to coordinate the needs of a engineering community with the development being done by a community of computer scientists," said Doru Marcusiu, who leads NCSA's Grid and Security Technologies Division and the effort to deploy NEESgrid at participating equipment sites around the country.

"In the past," Marcusiu said, "the conventional approach has been that the computer science community develops something, and the engineering community tries to use it, and it either works, it doesn't, or it can be improved."

However, NEESgrid's approach has been different in that the SI team first tried to identify the earthquake engineering community's requirements and build the NEESgrid infrastructure accordingly, allowing the science to drive every aspect of the collaboration.

What evolved out of intense and frequent discussions between earthquake engineers and SI team members was an iterative process in which the SI team would develop some software and turn it over to an "early adopter" group of earthquake engineers to try it out. With the feedback from the early adopters, the SI team would make adjustments in successive releases that would approach the engineering community's requirements more closely.

"The ability to keep an open mind throughout the process has been very important," said Severance. "Each time we released new software we learned new things about it. It required patience on the part of both the developers and the earthquake engineers, but we all learned not to give up just because a particular software release wasn't perfect."

The Future Of NEESgrid

In March of 2004, the SI team and the UIUC civil engineering department partnered with Kajima Corporation, a leading Japanese civil engineering firm, to conduct the first trans-Pacific experiment using NEESgrid. Like the MOST experiment, the experiment with Kajima involved both computational and physical simulation. Recently, Spencer and other members of the NEESgrid SI Team traveled to Kobe, Japan to discuss how NEESgrid could work with Japan's Earthquake Defense (E-Defense), a project which involved construction of the world's largest earthquake simulator. Similar partnerships are in the works with technological institutions in China and in Europe.

The work of NCSA and the rest of the SI team has been completed, but NEESgrid's work is just beginning.