I began this year with high expectations for broader application of grid computing. At the same time, though, I shared some concerns about vendors who might seek to warm themselves in the glow of grids fire while actually promoting their own proprietary approaches.
A welcome reinforcement of standards-based grid technology comes with this mornings announcement of the formation of the Globus Consortium, a non-profit clearinghouse for matching requirements with resources to guide the development of open-source grid infrastructures.
Before this mornings announcement, I spoke with consortium president Greg Nawrocki, who stressed the importance of grids beyond the applications that are most often mentioned as candidates for highly parallel processing. SAP, for example, has sought to lower its customers server hardware costs with virtualization technologies using Suns N1 Grid offerings. Grid computings potential “has been proven both in academic circles and in enterprises as well,” Nawrocki emphasized.
Its important to make clear the difference between an enterprise-class grid application and the massive but much simpler tasks that get lots of mainstream media attention. For example, Ive seen the [email protected] project characterized as an example of grid computing, even though theres no dynamic interaction among the 5.3 million users computers. To quote the [email protected] Web site, “Fortunately, the data analysis task can be easily broken up into little pieces that can all be worked on separately and in parallel. None of the pieces depends on the other pieces.”
[email protected] meets one key test of grid computing: its heterogeneous, with a huge variety of CPU types and operating systems contributing processing cycles to the task. But thats about as far as it goes. [email protected] doesnt begin to address the issues of “resource monitoring, discovery, and management, plus security and file management” that the Globus Toolkit has been built to take on.
The further challenge now facing the Globus Consortium is to make grids more credible outside the domains of research and technical computing. “When you look at the commercial applications space, people are looking for stable, robust, prudent solutions that are typically offered by multiple vendors,” explained Peter ffoulkes, Director of Marketing for High Performance and Technical Computing at Sun Microsystems. “In the scientific and technical space, where people are solving problems that have not been solved before, theyre willing to take on a little more risk to get the problem solved. Grids are perceived as having been brought to market to meet the needs of the risk-tolerant community,” said ffoulkes when we spoke last week prior to the public announcement of the Globus Consortium launch—in which Sun joins IBM, Hewlett-Packard, and Intel as a sponsor-level member.
Its no longer surprising that large numbers of computers can be harnessed together to work on a single problem. What remains to be proved to enterprise IT builders is that this can be done with dynamic allocation of heterogeneous resources in a demand-driven manner that actually saves money without compromising line-of-business dependability. Formation of the Globus Consortium sends a useful message that those needs are being identified and addressed with the support of major industry players—but in a non-proprietary way.
Theres only one week left before the entry deadline for eWEEKs Fifth Annual Excellence Awards. Information at www.excellenceawardsonline.com.
Tell me what you define as excellence in enterprise computing at [email protected]