Computers are an important part of the drug discovery process at pharmaceutical companies such as Eli Lilly. Typically, researchers make use of powerful Linux or UNIX clusters to do extensive and complex analysis. These clusters are in high demand and researchers often have to wait for them to become available. This has significant business implications, since Lilly estimates that delays during a major drug development program add to the total cost at a rate of as much as $150 per second.
In mid-2008, a Lilly researcher calculated that he needed 50 servers to do his analysis. The IT organization told him it would take 3 months to provide them. That very afternoon, he and Dave Powers, a Lilly IT specialist assigned to support the drug discovery unit, started experimenting with using Amazon’s EC2 service. Though they initially encountered a series of problems, they were able to iterate quickly toward a solution, creating a Hadoop cluster designed for data-intensive applications.
Much to their surprise, they soon realized they did not need 50 servers but 250. Working with virtual servers in the cloud, though, it was as easy as changing the number in a command line.
But suppose they had waited 3 months for their 50 servers and then found they were off by a factor of 5? This story is a good example of how speed and flexibility are often more important than price. Of course, this does not mean that Lilly will do all its computing in the cloud. Many of its systems run 24 hours a day, seven days a week, all year round. There is no particular advantage in moving them to the cloud at the moment, given current pricing.
Additionally, Lilly doesn’t want to put its most sensitive information in the cloud. Instead, it plans to build an Amazon-compatible private cloud inside its firewall, using Eucalyptus/Ubuntu software to assess where it could benefit from an internal cloud approach. It is also considering an SAP experimental sandbox.
The company recognizes that having the technical ability to exploit the cloud isn’t enough; management also must understand cloud computing’s possibilities.
From "Digital Trust In The Cloud," by CSC's Ron Knode