The chairman of the Global Grid Forum Mark Linesch last week opened the GridWorld conference in Boston spelling out the promise of grid. The new world is dynamic vs. static, shared vs. trapped in silos, virtual vs. dedicated, automated vs. manual and oriented around services vs. applications.
While academia is still the foundation for grid, Linesch says major grid deployments are happening. On the implementation continuum from early deployments to broad adoption, he estimates corporations are approaching the middle stage where momentum starts to build.
The drivers are varied but include getting more out of infrastructure investments, saving money and the need for higher performance. Not surprisingly, the early proponents are in industries that have the toughest problems to solve, such as the financial and life science sectors.
Robert Cohen, a fellow at the Economic Strategy Institute, says it takes pharmaceutical companies 20 years and costs $1 billion to develop a new drug, in part because 80% of new drugs go through clinical trials.
Companies can dramatically reduce costs by using grids to do more analysis in the virtual world, with the goal being to conduct trials on only 50% of the most promising drugs, and ultimately maybe only 20%. Such work is immensely compute intensive, and grids are proving to be the answer.
Cohen estimates that pharma’s adoption of grids will collectively result in 15% productivity gains and a 10% reduction in costs from 2003 to 2010.
Johnson & Johnson is one company already reaping rewards. Jeffrey Mathers, director of strategy and delivery in Pharmaceutical R&D Information Management, says the grid his group deployed last December now consists of 1,000 CPUs (half Windows, half Linux), supports six production applications, six regional locations and four business units involved in everything from drug discovery to clinical modeling and simulation. On any given month 16,000 to 20,000 GHz hours of processing is consumed.
“We have delivered good performance and proved grids are good for cost avoidance, but what I’m most proud of are comments from researchers who say we couldn’t do the research we’re doing today if the grid wasn’t there,” Mathers says.
In the financial sector, Cohen says some banks have found they can move $500 million from credit reserves to operating capital because grids enable them to better forecast credit risk. Being able to run the hugely complex jobs faster means they can run them more often and get better insight into current exposures.
Hurdles to broad adoption include software licensing restrictions, questions about security and cultural issues about sharing resources, but the demonstrable benefits make grid hard to ignore.
