By Kevin Davies
November 10, 2009 | It was last year when Amylin CIO Steve Phillpott knew he had a space problem. With his two San Diego datacenters perilously close to capacity, housing close to 700 servers and about 200 Terabytes of data, he was contemplating building a third data center in Las Vegas or Phoenix. But the notion of enabling a virtual organization appealed to the IT industry veteran, hence he chose to leverage the cloud.
Unlike most other big pharma companies, which might have one specific application in the cloud, Phillpott and colleagues took a broader view. “We wanted to build capabilities in several different areas,” says Phillpott. As the biotech’s business needs evolve, Phillpott’s staff can decide on the best tool for the particular business need.
Over the past year, Phillpott has built up Virtual Data Center capabilities in four major areas using Amazon EC2 and other providers. First is a pure infrastructure for quick provisioning (adding and subtracting) servers. “We actually have connectivity to Amazon,” says Phillpott. “Amazon is just an extension of my data center here. I can add applications up there.” Phillpott currently has four business apps running on EC2. “The beauty of it is I have secure connectivity so it looks like my data center.”
Second is software development tools, or what Phillpott calls “the Lego building blocks of building an application.” Often he doesn’t want to build an application from scratch, but rather pick apps off the shelf, such as Google Apps and Salesforce.com platform (Force.com).
The third area is support for disaster recovery—building capabilities to maintain images of Amylin’s servers and data using a cloud storage company called Nirvanix. “Now I’ve got a snapshot of every server in the organization stored within Nirvanix’ datacenter, such that if something was to happen to one of those servers, I could quickly rebuild it someplace else, like Amazon!” (Nirvanix competes with Amazon’s S3.)
The fourth area is the actual application of software-as-a-service (SaaS). Phillpott considers what applications are sitting on Amylin’s systems where he doesn’t want to support the hardware and software going forward. “All I want to do is rent to be a more variable model. If I add 100 people, I add 100 people times X dollars a month.” For example, Phillpott is migrating Amylin’s email solution, and considering some other large applications so that they’d be provided by SaaS.
A lot of companies have shied away from the cloud citing security concerns. Phillpott took the opposite tack: “The cloud’s here, it’s going to be here to stay, and I’m looking for applications that make sense out there.” Not every application fits the cloud, he says, but he’s already proven that a lot of apps do fit. “We’re starting to build maturity, capability, and a knowledge base, knowing that as we build that out, the industry matures and I can then keep moving more and more.”
The first Amylin apps running on EC2 are internal personal apps outside the R&D space. “We wanted to build a very solid security model before we tackled some of the R&D use cases,” says Phillpott. Those early examples include commercial sales (territory management), human resources (performance management), and some security apps. The savings are typically about 50% he says. Speed and performance considerations will inevitably become more important as Amylin runs more R&D apps.
Phillpott provisioned the first server at Amazon personally. It took him all of 15 minutes. “Now I haven’t provisioned a server in 10 years—that’s how easy it is!” he says. He showed his results to Amylin’s director of applications, and from that point on, his team was determined to outperform the CIO.
Lately, for some of the more complex scheduling, Phillpott has been working with Cycle Computing (see p. 28) in the research and genomics space. Another partner is a San Diego firm called Cirrhus 9, which has been helping Amylin on a couple of other cloud aspects.
Asked to comment on EC2’s performance, Phillpott says simply, “we’re very pleased. It’s taken us a long time to get to this place, because we wanted a solid security model.” Indeed, Phillpott says the EC2 security is greater than in most corporate environments he’s seen, although it is still improving.
However, Amazon is not the only game in town. “There are lots of companies coming into this space right now, but I don’t think all of them will be around in three years,” says Phillpott. “We want to ensure that, as we’re in the learning stage, we’re learning with someone who we’re confident will be around.” In a few years, once he has more expertise and the market matures, then “you bet,” he’ll entertain other options.
Reframing the IT Mindset
Aside from the obvious security questions, Phillpott says another key issue surrounds the mindset of the IT group, which worried that the cloud would limit or reduce career opportunities. “People at first thought, if we move everything out to the cloud, how will that affect me?” Phillpott recalls. “Over the last year, we’ve really educated them that, ‘No, it changes your skill set, it doesn’t do away with your job. I still need you to do things, it’s just that I need you to do higher-value tasks, focused on innovation and core business competencies, rather than commodity, keeping-the-lights-on activities.’”
Phillpott cites the example of applications being built up on Force.com. “The developers initially thought Force.com would replace the need for internal development. Once they tried building their first Force.com application, they said, ‘Oh, this just makes us more productive.’” Reframing is a real issue, like recoding a workflow application, he says.
Another issue is performance and latency. “For the most part, we’ve been setting up good solid connectivity with Amazon. I haven’t had to send massive datasets, but I’m identifying a couple of other applications that require sending large amounts of data in a couple of months. That will really test the connectivity.”
Like many others, Phillpott is also looking at internal virtualization and private clouds using VMware. This would be a complement to EC2, he says. “My goal is to have capabilities in these different areas, so when the business comes, I can decide: What is the best tool for that particular job? I don’t want to be forced into having a hammer and everything’s a nail—I have EC2 so everything has to fit there. I think we’re in a nice position where we have half-a-dozen different tools, and I can do the best selection based on cost, control, performance and security, as to which is the best fit.”
This article also appeared in the November-December 2009 issue of Bio-IT World Magazine.
Subscriptions are free for qualifying individuals. Apply today.