Sep 12, 2007
More than just computing power: Earth science on EGEE
You have a good chance of meeting Monique Petitdidier at any event that combines Earth science with computing. Although officially retired, Petitdidier continues in her role as a senior scientist with the French institutes IPSL and CNRS: she coordinates the Earth science contingent of the Enabling Grids for E-sciencE (EGEE) community, and is a driving force behind the DEGREE project (Dissemination and Exploitation of GRids in Earth sciencE). Petitdidier says that scientists in the Earth Science Research virtual organization view Grids as simply a computing means to their research end.
"Most institutes cannot afford intensive computation, such as that required to run millions of jobs or very large data archive exploration," Petitdidier points out. "This means that their science is limited by the computing power they have available."
"We mainly have researchers who need the Grid for specific problems, and will leave once they have obtained their results," she explains. "Sometimes they will come back with a different problem, but most of the time they just move on to other problems where you don't need a Grid. The emphasis really is on the science rather than the technology."
"Some applications need to share large data sets or algorithms, are using a statistical approach, and require rapid solutions. For example, if you have to reprocess a whole database with different parameters, this is a lot more adapted to run on the Grid."
But the benefits of Grids go beyond raw computing power, says Petitdidier.
"Grids are a great way to get people together and thus eliminate doubling of effort. We often see that people from different fields or regions have similar problems. For example, people looking at the flooding of rivers in France or Ukraine have been put in contact with scientists who have developed a similar application for the Danube river."
This doubling of effort can also be eliminated when developing software for geophysical sciences. Such software is often designed for a particular machine with specific characteristics, which makes it difficult to port to other architectures.
However, software developed for a Grid is ubiquitous, eliminating any problems with changing between machines, and leaving more time for pure research.
The Earth Science Research virtual organization in EGEE brings together some 50 scientists at any one time, but Petitdidier says she'd like to see more people get involved.
"We have many users but have not yet achieved a critical mass of institutes and organizations involved, so the challenge is to get new people on board," she says.
• This article was published online in iSGTW on 11 July.
About the author
Hannelore Hämmerle, EGEE