Ending the Grid Lock
Grid computing has long found a following in academia, the financial industry and pharmaceutical enterprises, but other industries have been slow to follow. Until now.
The continuing march of grid computing recently received a nudge forward when a consortium of big name IT companies formed a group specifically to promote the use of grids in the business world.
IBM Corp., Sun Microsystems, Hewlett-Packard and Intel have thrown their collective weight behind the Globus Consortium, a non-profit organization aimed at turning more businesses onto grid computing – giving companies in the auto, aerospace, and energy industries the open source tools to develop their own grid systems, and educating them about how to best use those systems.
Launched in January, the organization’s main goal is to make grid computing more accessible to corporate technology users. However, its roots stretch back several years to one open source project.
Originally released in 1997, the Globus Toolkit is a set of software applications that can be tweaked by users, which allows businesses to tailor the applications to fit their exact needs. Based on Unix, the toolkit offers a set of components that can manage security, information infrastructure, resource management, and communication across a (typically heterogeneous) grid.
The toolkit, funded by the Defense Advances Research Projects Agency (DARPA), was borne out of work from Steve Tuecke, Ian Foster and their colleagues at Argonne National Laboratory.
However, the software sparked more initial interest with niche groups looking to solve big problems. Over the years, grid computing has found more of a following – and received more media attention – for the flashier academic projects like SETI@Home and the Human Proteome Folding Project that reached out to average computer users to lend their unused computing power to trying to contact aliens or decode proteins. While this is one forum for grid computing, Tuecke says that “cycle scavenging” projects like these only represent the most simplistic view of grid computing.
Now, though, more traditional sectors have started to adopt the process. Financial service and pharmaceutical companies have already embraced grid computing – sometimes at a basic level, borrowing computing power from under-used desktops at their companies and other times on a more complex basis such as running applications run in a loosely coupled environment.
These companies use grids to run the computationally intense “Monte Carlo simulations” – a type of probability analysis based on statistical sampling – necessary to analyze portfolio holdings or derivatives risks or to look at the countless permutations of molecules that might be combined to create new drugs.
But other business sectors have been slower to join the computer-sharing smorgasbord, reportedly concerned about the potential security risks of working in such a diverse and decentralized environment, and the relative lack of resource monitoring and management software and services that have only recently started to become available.
The consortium hopes to win over those holdouts, promoting the products and services of its own members and also offering its own free open source toolkit to aid in development efforts.
Given support as well as better hardware and applications from the likes of Sun and IBM, Tuecke and Nawrocki both believe security- and sophistication-conscious businesses will become more open to grid computing as a more efficient means to solve business problems without investing in more expensive supercomputers.
Telecommunications firms are already starting to wake up to grid for supporting and load balancing their next-generation networks, says Mark Linesch, chair of the standards-setting Global Grid Forum.
And Linesch says manufacturing, aerospace and oil and gas companies will soon embrace grid computing in greater numbers, an idea Insight Research of Boonton, N.J. backs up with estimates that grid computing-related expenditures to jump from $714.9 million in 2005 to $19.2 billion in 2010 in telecommunications alone.
Despite solid academic pedigree and references from some of the high-tech industry’s biggest vendors, the Globus Consortim may hit some bumps on the road to wider commercial acceptance.
There is at least one potential standing in the way that has done a good job creating fear, uncertainty, and doubt about open source business solutions.
Just as Linux application developers have found Microsoft roadblocks at every turn, it looks as though open source grid computing may face competition from the Redmond behemoth as well.
Ironically, Microsoft provided “substantial funding” to Globus as recently as 2003, says Tuecke, and the software giant was part of the Globus Alliance – predecessors to the Globus Consortium and promoters of the toolkit.
Microsoft was noticeably absent from the consortium, and is rumored to be working on its own as yet unannounced set of grid computing software tools under a project reportedly code-named Bigtop.
Without commenting on specific plans, Microsoft issued a statement instead that says the company “doesn’t have any specific grid computing plans to discuss at this time While there is much interest in grid, there’s still much research and investigation needed to ensure the technology is actually focused on addressing important business problems.”
As for Tuecke, he doesn’t deny there may be an opportunity for Microsoft to upset the apple cart of Globus’s open source development – if it can act quickly enough.
“You can’t dismiss Microsoft, they’ll make a difference if they decide to make a difference,” he says. “But in some ways, the train is already leaving the station.”
Regardless of the Microsoft threat, Nawrocki admits that grid computing “needs to get a little more industry hardened, first and foremost there’s still a lot of outstanding bugs.”
Further, he says that a lot of popular business applications haven’t been adapted to accommodate the decentralized architecture of grid, and there’s still a relative lack of tools that would make monitoring the decentralized and far-flung work on the grid easier to track.
Nawrocki believes, however, that much as a ready application community eventually sprung up around Linux, a cadre of new grid application developers will emerge to help shepherd enterprises on grids.
“I really think, at this point, it has to do with an application disconnect,” Nawrocki says. “Many application developers don’t know what grid is yet.”
If the Globus Consortium succeeds, though, that won’t be a problem much longer.
Become an MIT Technology Review Insider for in-depth analysis and unparalleled perspective.Subscribe today