Grid's promise remains strong, but it is turning out to be more difficult than expected to develop the infrastructure for Grid computing, Fran Berman, director of the San Diego Supercomputer Center and the National Partnership for Advanced Computational Infrastructure, said in a keynote speech to Global Grid Forum 5 in Edinburgh last week.
"The promise of the Grid has been not been oversold, but the difficulty of developing the requisite Grid infrastructure has been underestimated," Berman said. "There are many questions not addressed by TeraGrid and other projects that must be addressed to develop a usable and useful Grid information infrastructure."
"We have barely scratched the surface" on issues such as: program development environments; debugging, compiling and performance tuning; fault tolerance; modeling of dynamic, unpredictable environments; Grid market economy (allocation, accounting, cost models); new Grid programming paradigms; and "extreme heterogeneity" (sensors, supercomputers, cell phones, cars, etc.), she said.
The ultimate goal is "a useful, usable, stable Grid" that is rich in resources and options, able to adapt to new technologies and uses, is scalable, and is more cooperative than competitive, Berman said.
Applications are the key to Grid's success, she said. The goal for Grid infrastructure is "to some day be as natural a part of the picture as the OS," Berman said. "Grid will be considered 'oversold' if the only people who can productively use it are the techies."
Next generation Grid applications should be adaptive (run where you can find resources satisfying criteria X); real-time (do something right now); coordinated (dynamic programming, branch and bound), and "poly-applications" (choice of resources for different components), she said.
"We still can't throw any application at the Grid and have software determine where and how it will run," she said.
Berman discussed one possible model for future Grid applications: Everyware, a highly adaptive Grid application which investigated solutions to the Ramsey Number Problem. Everyware was not embarrassingly parallel, but used branch and bound and simulated annealing.
The GrADS (Grid Application Development Software) Project is working on design and development of a Grid program development and execution environment, she said.
The explosion of data could benefit Grid, she said. "Over the next decade, data will come from everywhere," she said, such as scientific instruments and experiments, sensors and sensornets, and new devices such as personal digital devices, computer-enabled clothing and cars. That data will be used by everyone, including scientists, consumers, educators and the general public.
"The software environment will need to support unprecedented diversity, globalization, integration, scale, and use," Berman said. Among the challenges for Grid technology is to "combine data, knowledge and information management with simulation and modeling."
Current real-world distributed applications, such as Wal-Mart's real-time satellite inventory control system and Everquest's online gaming, "demonstrate that it is technically, commercially, and economically viable to deploy robust, large-scale distributed applications," Berman said. "The Grid should accelerate that progress."
"The availability of Grid services should allow designers to build on existing infrastructure and evolving technologies," she said, and applications developed for the Grid will in turn contribute to community infrastructure and standards. "Grid applications will help evolve policies for a scalable Grid," she said.
Images
IE 7
Firefox 2.0
