InformationWeek 500: Eli Lilly Ties Future To Cloud
The pharmaceutical firm relies on the public cloud for compute-intensive scientific and drug development research and sees more extensive use of cloud computing as essential to its future.
Eli Lilly CIO Michael Heim talked about his firm's implementation of cloud computing for the first time at the InformationWeek 500 Conference in Dana Point, Calif. today. He said his company has found value in using Amazon's EC2 for bursts of research computing power. The cloud is now firmly embedded in Eli Lilly's approach to IT for the future and the company is exploring longer term uses in addition to offloading short term research jobs.
Heim and Michael Meadows, Eli Lilly's vice president and information officer, global information services, addressed InformationWeek 500 Conference attendees. They said their firm was banking on continuing to spend 20% of revenues on R&D and stretching that spend by exploiting the cost savings of cloud computing. Heim said data security would always be an issue, but Eli Lilly was able to send early research data into the cloud, combine it with other publicly available data such as genome research and execute a compute intensive analysis on Amazon's EC2 infrastructure.
The data being analyzed is "pre-regulation" data, said Heim. Analysis of clinical data assessing test results on particular pharmaceutical compounds is still performed in-house under more secure computing methods.
"We think the cost advantage is real and the complexity issue can be managed," Heim told the 370 attendees at the event. Eli Lilly made its first foray into cloud computing as its scientists sought quick and cheap horsepower with which to execute their studies. They often needed 800-900 CPUs for 850 megabytes of data over several hours to generate simulations or models of chemical interactions.
"Often it's a scientist at the workbench doing research who is taking the lead in moving our understanding of cloud computing forward," said Heim. But Eli Lilly IT has created 15 templates that make it easier for scientists to commission research servers in the cloud and authorizes scientists to self provision them with a minimum of IT interference.
"The cost is trivial in some cases for what they're doing. It's hard to overestimate the value of letting scientists work at their own pace," said Heim.
He was seconded by Meadows, who said, "The cost model is amazing in what it allows in procurement under defined limits of corporate credit card use," he said. EC2 users with an account can enter their credit card number and commission a workload at a rate of 8.5 cents an hour on a simple Linux server, 12 cents an hour on a Windows Server, or 22 cents an hour on a Red Hat Linux Server. Virtual servers with more memory and CPUs are available at higher rates.
Meadows said Eli Lilly has been using cloud computing for newly commissioned research workloads and shipped them off to the public cloud rather than buy more servers for the data center, a practice often referred to as "cloud bursting." But he said the firm was encouraged by that practice to look at longer term uses, where servers might be lined up for a certain number of hours of use over a year.
"I thought the extreme affordability was only in that temporary cloud bursting. But we've been costing it out over a 3-5 year period and there's benefit there as well," Meadows said. Just because a company has a long-term contract with a cloud vendor doesn't mean that it commissions servers and leaves them running 24 hours a day, as it would in its own data center. It can contract for hours and run jobs for a certain number of hours a day when they're needed, he said. One issue that's hard to resolve is the cost of moving large data sets into the cloud, for which there is a bandwidth charge. Frequently the data set becomes larger as other data is added to it. The user then faces a dilemma: moving data out of the cloud is more expensive than moving it in, leaving some to conclude that one can end up trapped in cloud expenses. "We're still evaluating the cost of data flow back and forth," said Meadows.
But Heim reaffirmed that cloud computing has been asset for Eli Lilly's R&D. The firm spends 20% of revenues on developing and evaluating clinical tests that may result in a successful new drug; bringing a new drug to market now costs as much as $1 billion per drug, he said. Patents on some of its existing moneymakers are due to expire soon, and generic lower cost substitutes will appear. Other pharmaceutical firms were growing by acquisition, but Eli Lilly will keep investing in its own research, Heim said. "When revenue is declining, it's important we do everything we can to reduce the cost of the development phase. Anything we can do to bring down costs will help us put more candidate pharmaceuticals in the pipeline," he pipeline.
Prior to adopting cloud computing, Eli Lilly had filled its two data centers and "we didn't want to build another." Going out into the cloud, combined with server virtualization in-house, has eliminated the prospect that they would have to build a third center.
Eli Lilly recently made headlines when it was reported in a dispute with Amazon on SearchCloudComputing.com over service level agreements available on EC2. Eli Lilly was allegedly seeking greater assurances of server availability, and Amazon was refusing. Amazon CTO Werner Vogels eventually issued a tweet that said Eli Lilly remained a customer and no ongoing dispute separated the two parties.
Asked about that report, Heim resorted to citing the locker room dialogue between a player and a reformed sports agent, played by Tom Cruise, in the movie, Jerry McGuire. What sounded like an argument to bystanders was a way of communicating between two strong-willed parties. The exchange ends up: "I think we're finally talking."
Heim said Eli Lilly was concerned about being locked in to any cloud vendor, but any lock-in dangers were offset by the benefits. Meadows mentioned discussions with three cloud vendors, Savvis, BlueLock and Amazon. But Eli Lilly is currently engaged in cloud computing only with Amazon, and he was vague about the nature of any future business with Savvis and BlueLock.
An immediate future possibility is collaboration in the cloud with other pharmaceutical researchers. Eli Lilly is exploring the possibility of a network of collaborative firms but stopped short of predicting an industry-specific form of cloud computing might emerge.
Nevertheless, Heim says cloud computing is essential to Eli Lilly's future. It's competitiveness relies on unleashing the creative research powers of its scientists and producing new products. IT's role must change in the face of the changing needs of the business. "We're trying to change the IT mindset from saying 'no' to the right way to say 'yes.'" FURTHER READING:
InformationWeek 500: Innovation Spurs Business Growth
InformationWeek 500 Preview: The Growth Imperative
About the Author
You May Also Like