SCIENCE

Gather a bunch of scientists together and they’re bound to talk to each other. It’s not quite so easy with the computer systems they use. And, as science becomes increasingly multi-disciplined, that’s a stumbling block. Predicting severe storms, for example, relies heavily on computers. But the computers have different tasks and different programs: mining data and integrating data with real-time instruments and simulations. Together, these tasks will eventually drive online instruments such as Doppler radars, but only if the tasks can work efficiently in concert. A team of IBM researchers has published an approach that allows that flow. Their work, “Building Web Services for Scientific Grid Applications,” appears in the latest online edition of the IBM Journal of Research and Development. The key is assembling a “grid,” an amalgamation of the resources used to host each scientific team's software applications along with software services that tie them together. But the definition is a lot simpler than any implementation. There are security issues – how does a one scientist gain access to another’s data, for example – and the scientific software lacks the graphical user interface (GUI) so familiar to PC users. The IBM team presents solutions for these and other challenges. The framework they describe allows scientists to “wrap” their software applications as web services and deploy them on a grid; it also automatically provides an authorization system that allows scientists to compose scientific workflows using these services and to monitor the status of their workflows on the grid.