Harvard University and IBM Corp. are developing a universitywide computing grid for student and faculty research, data sharing and collaboration in life sciences, engineering and applied sciences, they announced Wednesday.
The "Crimson Grid" will be based on OGSA (Open Grid Services Architecture) and is expected to eventually be available to other universities in the region, said Robert Eades, IBM's worldwide executive for academic, government and health in the company's life sciences division. The grid also will be part of a Massachusetts biotechnology grid.
"I am just excited beyond words," said Joy Sircar, chief information officer and IT director at Harvard's Division of Engineering and Applied Sciences. His division is where the grid will start and then it will be expanded to the rest of the university and beyond. The grid is a milestone in Sircar's IT career and he said that it has the potential to be a transformative step in how science and engineering tools and technologies are developed because of Harvard's reach and importance in those fields.
Harvard and IBM will develop and pre-test tools and protocols for the grid. Harvard is receiving an IBM Shared University Research award as part of the initiative and will receive e-Server systems for a blade center that will power the grid. Harvard's Division of Engineering and Applied Sciences' IT group, along with IBM computer scientists, will implement and build the Grid Reference System Implementation, which is the grid's core development environment. The Faculty of Arts and Sciences and Harvard University Information Systems will provide the network backbone service.
The reference implementation can then be shared with other universities, agencies or groups interested in developing a grid, but that perhaps lack the time or expertise to develop such a computing system, Eades said.
"The really key benefit (of the project) is going to be having an easy-to-use, easy-to-deploy grid reference implementation," he said. Often, customers will take such an implementation to develop a grid and then buy commercial products for use in the grid -- in the case of life sciences that could be relational database software, he said.
"A very, very early test (implementation) is already up and running," Sircar said. The grid will be "an evolving process" over the next 12 months or so, he said.
"Developing the grid involves a lot of research, a lot of integration," he said. Integration of platforms, compilers, tools, resources and different nodes has to be accomplished. And that is just at Harvard, which has its main campus in Cambridge, Massachusetts, but offices and property all over the Boston area. Over time, the grid's reach will be expanded and it has the potential to include sites nationally and internationally because a lot of research done by Harvard students and faculty involves collaborations with other universities, research institutions and agencies, Sircar said.
Grids have the potential to alter how IT is approached, leading to more collaboration and a different view of the "research computing IT ecosystem," in Sircar's opinion. Social sciences and applied sciences could begin to affect technology design as more collaborative computing takes root and becomes the norm, he said.
"Why rely on Bill Gates to give you technology and say, 'This is what is good for you to use,'" he said. Instead, scientists and researchers will increasingly insist on letting vendors know what commercial products they want and need. As things are now, many scientists and researchers develop their own proprietary software and tools because they don't like what is available commercially or what they need doesn't exist. But that could, and likely will, change with more integrated, collaborative efforts, such as the grid concept, Sircar said.
Grids are an emerging networked computing method particularly useful in areas such as sciences and engineering that are compute intensive, where massive amounts of data are accessed and analyzed. The commonly used analogy is to electrical utilities, where power is switched on only when it is needed. Although a grid system might be complex, involving many machines in many locations, the user is meant to "see" just a single virtual environment, akin to how the Internet works with servers contacting other servers each time a user accesses a Web page. All of that happens in the background. Grids also can tap the computing resources of machines that are not otherwise in use at that moment, taking advantage of all of the power that is available in the system.
"We tend make it sound like it's all one grid," Eades said. "It's really grids of grids."
The Crimson Grid is expected to enable collaborative research by faculty and students across disciplines, as well as data sharing. "Crimson" is the nickname of Harvard athletic teams, the name of the school newspaper and the color most associated with the university.
