By Robert PursellDecember 15, 2006
(SUFFOLK, Va. - Dec. 14, 2006) -- The High Performance Computing Modernization Program (HPCMP) recently assigned a super computer to U.S. Joint Forces Command (USJFCOM) that will enhance experimentation and training efforts in modeling and simulation (M&S).
The super computer itself is much larger and more powerful than the machines used today which will yield finer details when it comes to imaging and behavior at a faster speed.
The super computer will be operated mostly by the Joint Training Directorate (J7) and Joint Experimentation Directorate (J9), housed in the Joint Training and Experimentation Center (JTEC) and accessed through the Defense Research and Engineering Network (DREN).
The DREN is an official DoD network specifically designed for computational research, engineering, and testing, and is used to transfer leading network and security technologies and capabilities across the DoD and other federal agencies.
Jim Blank, USJFCOM J9 modeling and simulation division chief, explained the command's focus in using the machine.
"There's been a shift in focus, as you can imagine, from rolling deserts and plains to an urban environment," he said. "You can't model an urban environment without modeling the people. That is the most important part of the city. Secondarily, is the complex terrain involved in that environment. The super computer helps out in both areas."
Tony Cerri, J9's experimentation engineering department head, gave an example of how the super computer will affect a simulation of Baghdad.
"In a city like Baghdad, we can say this would be morning rush hour, all of the sudden 500,000 people get up and go to work. That's not something that we've been able to do very well," said Cerri.
"It also allows us to do some experimentation with other ways to use supercomputing, but that's work to come."
Blank discussed the difference between the horsepower of a regular computer versus the super computer and how it impacts each individual item (called an "entity") in these simulations.
"It's fidelity versus scale. Typically, as you've increased the number of entities that you put into a simulation, your resolution of any particular entity has gone down because you just can't support a million entities at a constant level of resolution. Our entities have behaviors associated with them. Now we can maintain the full behavior characteristics of the entity as we scale out to a million," said Blank.
"In a previous life, we ran about 32,000 entities at any given time. That was probably the max that we were capable of. With super computers, you can run over one million entities, and we've done it," he said.
Blank summed up the advantage of having the super computer housed at USJFCOM when he said that having the machine locally will enhance capability and make development much easier.
USJFCOM did not accomplish this effort by itself. The University of Southern California (USC) Information Science Institute (ISI) played a major part in helping the command work with the HPCMP to acquire the super computer. USJFCOM was negotiating for about a year before it received the final approval.
"They have significant super computer experience and we worked fairly close with them because of their expertise to keep us smart, engaged, and in the right direction," said Blank."