Lilly Looks to Cloud Computing

Jan. 6, 2010
An interview with Eli Lilly Systems Engineer Dave Powers on his company’s use of cloud computing.

PhM: While many people are still unfamiliar with cloud computing, Lilly is already using it to support its scientists. Can you share how it’s being used and to what extent?

D.P.: If we look at the common 3 layer model of cloud computing (IaaS, PaaS, SaaS), our primary focus and energy has been spent leveraging and building on top of the IaaS capabilities of Amazon’s AWS offerings. We have focused on our Discovery Research innovation engine and have taken advantage of the flexibility and agility EC2 offers for our High Performance Computing (HPC) algorithms. In addition, we are also taking advantage of the fact that we can leverage AWS to enable self-service capabilities internally which dramatically reduce cycle time from idea to execution. We are currently on target to have up to 10 HPC applications “cloud enabled” by end of the year and to have release 1.0 of our self service utility computing offering rolled out internally that we affectionately refer to as our Vending Machine Computing environment.

PhM: In that cloud allows for a sort of virtual, web-based computing, how do you see it complementing, or replacing, work performed on your internal servers and platforms?

D.P.: We continue to aggressively move forward internally with virtualization of existing hardware assets in our data centers which allows us to shrink the footprint in our data centers in terms of the numbers and quantities of server and storage platforms. The approach that we started out with a couple of years ago was to move to virtualized environments. What we have seen with Amazon’s cloud capabilities is that virtualization is not the end game. Rather it’s an enabler of new innovation and new capabilities that we can leverage to enable the research engine of Lilly. With the recent introduction of Amazon’s Virtual Private Cloud (VPC) we see that as a next logical step for the enterprise to leverage as we move from “owning it all” in our data center to leveraging infrastructure on-demand that is not completely exposed to the public internet.

PhM: Cloud computing gives you access to high-performance computing—is the need mainly in R&D, or might there be a need in manufacturing or other parts of the corporation?

D.P.: The High Performance Computing space was a natural fit for us on day one when we started looking at cloud computing offerings as it is where we saw instantaneous benefit from the AWS offering. It was on-demand, scalable, elastic, had cost transparency like none we had seen and was very low friction to start using immediately. We used HPC as our beach head to test and build confidence and to build on the experiences and knowledge gained. While advanced analytics is a strategic direction for Lilly going forward and cloud computing will play a significant role in enabling those analytics, we also very much see the cloud as the point of integration for external (and internal) collaborations. In addition, as we gain confidence and momentum, we are also looking higher up in the cloud stack and will be leveraging PaaS from the Microsoft Azure offering and SaaS from a host of vendors as we collaborate externally more and more.

PhM: Manufacturers like Lilly are using vastly increased metrics and analytics, often in real time, to monitor plant operations and assure quality. Could cloud computing facilitate these efforts?

D.P.: Just to be clear, we are not of the mindset that everything is ready to move to the cloud today. That is just not the case. Neither are cloud vendors ready nor are enterprise applications ready for such a bold move. Rather, we are being very thoughtful about which applications and/or data are cloud ready and which are not. To be sure there are manufacturing systems and real time systems that are absolutely critical to our day-to-day business and any disruption to those process workflows could prove a significant impact. I think it’s safe to assume that these types of environments are not going to be at the top of our list of candidates to start moving to the cloud on day one.

In addition, we also must consider that we will have a multi-cloud approach. That is, we will have a private cloud internally, the public cloud, and a hybrid cloud. So, when you ask if cloud computing can facilitate these efforts, we would need to distinguish between which “cloud” might be used and the classification of system and data to be moved to the cloud. My guess is that manufacturing systems will not be part of our early adopters of cloud computing.

PhM: Cloud could also, as you’ve said, be a “point of integration” between Lilly and outside researchers? What kinds of experimentation are taking place in this area?

D.P.: We are working on some small prototypes of collaboration with external parties right now including sharing of algorithms for analytics and securely exchanging data with collaborators. The bottom line is that until we can figure out the whole Identity Management issue, my vision of leveraging the cloud as a point of integration will lag behind. We are also working on some creative ways to collaborate with external partners in the cloud, but it involves additional work and overhead that could be done away with by solving the Identity issues that hinder a more free flowing collaboration environment.

PhM: What are the data security issues that you’re confronting with cloud, especially in working with external scientists or other companies?

D.P.: We have the same data security issues that everyone else has faced including: encryption of data in transit and at rest, limiting access to machines by using keys only,  multi-tenant exposure, increased threat vectors from a large shared environment on the public internet and host of others. I think what cloud computing has done for us is to show us that the days of living with a feeling of comfort and security behind the firewall on our corporate network are starting to come to an end. Which has the effect of changing how we approach security in our own software development practices internally as well as what we expect from our COTS packages and vendors. We have worked closely with our privacy and legal folks over the past year to engage them and seek their guidance and counsel on collaborations with external partners using the public cloud.

The implication of some of these security issues in the public cloud is that it has a limiting factor in terms of how fast we can move with partners, what kinds of data we can exchange with them and ultimately how much value we get from the cloud capabilities.

PhM: Finally, how about a glimpse at the future: Where will cloud computing have the greatest impact upon Lilly and its operations in the future?

D.P.: Going forward, I see cloud computing expanding its role in our advanced analytics strategy. In addition, based on the cloud ecosystem’s innovation engine that has moved into full swing in 2009, I expect to see many of the Identity and Security issues to be solved by really smart and creative people which will truly enable us to take advantage of the cloud as our point of integration for collaboration with partners globally.

I believe the cloud will also start to steadily become the foundation of our computing environment going forward and I would expect the enterprise datacenter profile to change from an “own it all” model to much more of a “utility” based model leveraging internal and external cloud capabilities. Beyond IaaS which has its obvious immediate benefits to us today, we are already talking about expanding our movement into the higher layers of the cloud stack including PaaS with Microsoft Azure and SaaS for a number of applications that we see expanding around the world with partners and collaborators. 

About the Author

Paul Thomas | Senior Editor