PPPL developes internet-based simulation capabilities
Click here for a high resolution photograph. |
Two years ago Raffi Nazikian, deputy head of the Princeton Plasma Physics Laboratory (PPPL) Off-site Research Department, worried about traffic jams. He wasn't concerned about the kind of bottlenecks that develop along U.S. Route 1, just outside the Laboratory in Plainsboro, NJ. Nazikian was concerned about roadblocks to the accessibility of plasma simulation tools by fusion researchers. Plasma is the hot, ionized gas that is used as a fuel for the production of fusion energy. Physicists use powerful computer codes to interpret existing data and to predict the behavior of plasma confined by a magnetic field in fusion energy experiments.
Nazikian observed that the users of these computer codes were nearly always the software's developers themselves. "To put this problem into context, if the developers of Microsoft Word were the only folks who knew how to use it, there might be no more than one hundred people in the world running Word. If you wanted to run it, you would have to be a software engineer," said Nazikian. He expressed his concern to Doug McCune of PPPL's Computational Plasma Physics Group (CPPG). As a result of this interaction and subsequent efforts of the CPPG, an unprecedented improvement in accessibility and use of plasma codes is on the horizon.
Currently, if researchers want to utilize a plasma simulation code, they usually seek the assistance of a computational specialist with the detailed knowledge required to access and run the program. But the code experts are often overloaded with demand, resulting in a huge backlog. "So we ration the applications of these tools, resulting in major bottlenecks in the interpretation and understanding of experimental data and the planning for future experiments," said Nazikian.
It's not surprising that Nazikian and his associates, who share the world's fusion facilities and software remotely, would be the first to realize this pressing problem. "On a small scale, we are dealing with the very same issues that the broader fusion community is going to have to face on ITER, the large international fusion energy device that will be located in France. A very large number of scientists, located around the world, will work remotely on this one facility, conducting a relatively small number of experiments. To do this efficiently, they're going to have to design experiments in excruciating detail before they run them. Consequently, ITER researchers will have to simulate these experiments thoroughly beforehand, so they're going to need easy access and use of simulation codes," Nazikian explains.
Now, innovative work by McCune and fellow PPPL computational scientist Eliot Feibush may solve this problem well in advance of ITER by providing plasma physicists with web-based access to powerful computer codes. "Plasma physics has a lot of immensely complicated codes that are labor intensive to move around to different labs. There is considerable engineering cost to get these humongous programs to run correctly on different, incompatible computers. It would be great if folks who are not located at PPPL could use our simulation programs via the web, and if we could access their software, without us all having to download and install the large codes on our individual computers," notes McCune. The idea is to have physicists who are not necessarily code developers interact with shorter, simpler web-based interfaces that communicate with computer servers at various laboratories.
To this end Eliot Feibush is working on a web-based graphical user interface for the simulation of reflectometer measurements in plasmas. These simulations run codes developed by PPPL's Gerrit Kramer and Ernie Valeo. PPPL researchers and their collaborators worldwide build and install reflectometers as diagnostic tools that measure plasma turbulence. The reflectometer sends a radiofrequency wave field into the plasma, which is reflected at a specific location. The interference pattern that comes back to the reflectometer is indicative of the turbulence at that location.
Feibush's graphical interface allows users to access and input real experimental data, including various plasma parameters such as electron density and temperature and magnetic field strength into their simulations, yielding realistic results. Physicists will be able to test specific reflectometer designs before the devices are built, saving time and money. Web access will enable effective collaboration in reflectometry by permitting colleagues at other labs to utilize the advanced simulation capability developed and hosted at PPPL. Users will always have the latest version of the software at their disposal, so they will no longer have to worry about compatibility.
The graphical user interface program under development at PPPL is Java-based. Programs written in this language can run on either PCs or Macs, eliminating the need for distinct versions of software for each platform, such as required for Fortran, C++ and other compiled languages. Furthermore, Java has built-in graphical and user interface components, such as buttons and sliders. It was designed for the Internet and is currently used extensively for selling merchandise on the web.
But scientific applications are much more complicated than accessing an on-line catalog to buy a pair of shoes or clothing. According to Feibush, "After setting up the input parameters and starting a run, the interface software monitors a simulation that performs much more computation than typical commerce programs. Visualizations of the results enable the physicists to evaluate the effectiveness of simulated reflectometers. The challenge is to develop quality scientific graphics applications in Java that work over the web with existing physics codes."
As a result of Feibush's work, PPPL will soon unveil to the fusion community a robust web-browser-based user interface for the reflectometer code that comes packaged with sophisticated graphics capability. The interface will allow physicists to run simulations of complex, sophisticated experiments effectively without having to learn about the underlying code's file structures, subroutines, directories, platforms, and naming conventions. According to Nazikian, "We are now trying to come up with a standard for duplicating this interface tool for any plasma simulation code. We have an idea of how this is going to translate generically to make other codes available to the broad user community. So I think this is the beginning of a major sea change in our way of thinking about computational tools in the fusion community and how we use them."