Drug Companies Wont Swallow the Grid Pill

Although pharmaceutical companies need high-performance computing to do drug research, a new report finds that the vast majority are in the early stages at best when it comes to learning about grid computing.

Youd think that pharmaceutical companies would be ideal candidates for grid computing. They have a great need for high-performance computing in drug research, have vast amounts of data and are challenged by being one of the most geographically dispersed industries.

Youd be wrong, according to a new report from The 451 Group.

According to William Fellows, an analyst at the research firm and one of the authors of "Grid Computing: Adoption in the Pharmaceutical Sector," the perceived wisdom of researchers is that less than half of the top 20 pharmaceutical companies are implementing grid.

Of the ISVs that supply the sector, IBM claims to have relations with 50 percent of these companies. But as far as Fellows can discern through his research, the vast majority is "either at a very early stage or [does not have] what anybody would think of as an enterprise deployment," he said.

The researchers found that leaders such as Johnson & Johnson and Novartis AG are making great strides in grid adoption, moving from small grids to thousands of devices, gearing up to share infrastructure across their companies and putting enterprise IT applications onto grids.

But, the report says, for the majority of pharmaceutical companies, the jury is still out on grid, with many companies having failed to find a compelling reason why they should adopt the technology.

"Not one drugs been discovered using grid all the way through discovery through licensing to regulatory approval on the other end," Fellows said. "Once use cases start to happen, it will encourage momentum. But were not as bullish on wider adoption of grid in this sector as some boosters are."

After discussions with pharmaceuticals, The 451 Group reported that "the vast majority" either dont see grid as offering an advantage over the technology theyre already using or are using it in limited tests or pilot projects for research and development.

This doesnt make sense on a lot of fronts, Fellows said. For one thing, because of M&A (mergers and acquisitions) activity, the industry is more dispersed than any industry the research firm has looked at.

Another issue is federating and aggregating data—a problem thats common to many industries and explains why ISVs such as IBM and Oracle Corp. are working hard to promote their own federation and aggregation technologies.

"The top 10 pharmaceutical companies have this challenge where data from various programs is distributed all over the place," Fellows said. "They need to bring it together, integrate it and have it able to be [analyzed] without regard to original format or where it needs to be stored."

As such, the industry could well benefit from grid in the same ways that other sectors are benefiting: by using spare computing cycles, reducing hardware costs by hooking up underutilized boxes, increasing cooperation and communication, producing additional analysis, and keeping pace with competitors through more rapid product development.

Whats holding pharmaceuticals back? The same thing thats holding other industries back: software licensing, cultural organizational issues, data management issues, user resistance to running applications across the PCs of researchers, no recognition of a compelling application, and the perception that current distributed computing architectures are working just fine.

/zimages/1/28571.gifClick here to read about Macrovision and Platform teaming up to fight the problem of software licensing across grids.

Those companies may well have just cause to pause, Fellows said. For one thing, processing over wide areas without regard to where resources are located doesnt always pay off, since potential savings are offset by overhead.

"A couple [of firms interviewed] said, We cannot and arent in a position to implement wide-area grid because of the overhead associated with doing this," Fellows said. "Theyre geographically dispersed. They can take a job and drop it on a grid and have it spread over resources located around the world.

"But the time it takes for those tasks to go find the right database, go find the right data set and take that pretty significant chunk of data back to where the application is, to then have it processed and resynchronized—the overhead in time and money is too costly at the moment," he said. "So, they decided not to go ahead with wide-area deployment. That has to do with the cost of bandwidth."

At the heart of the resistance also may be the fact that pharmaceutical companies are managed by scientists who are more familiar with wet stuff—"test tubes with stuff in it," as Fellows put it. "Grid and other ways of doing modeling analysis havent taken hold within organizations in a fundamental way yet," he said.

The key to change, Fellows said, is to have organizational champions who understand the benefits of IT positioned within the pharmaceutical and other industries.

Its either that or luck. "Its having the infrastructure that has the right elements in place at the right time, so you can take advantage of certain desktops in certain areas and harvest the power," Fellows said.

/zimages/1/28571.gifCheck out eWEEK.coms for the latest database news, reviews and analysis.