The big problem with calculating the ROI of cloud projects is that IT departments aren't even able to do cost-based accounting for the services we already offer on premises. So how can we compare a new way of doing business? Trending our latest InformationWeek Cloud ROI Survey against the same poll done in April 2010 shows that we are, however, making some progress. For one thing, since we wrote our 2010 report, the definition of "cloud" has stabilized, and we're now convinced that public, private, and hybrid cloud computing models are here to stay, an opinion that's confirmed by our 2012 State of Cloud Survey.
In particular, half of respondents to our cloud ROI survey say they frequently or often need to quickly add computing capacity, up five points since 2010. For them, a hybrid model makes good business sense. DreamWorks Studios illustrates the "cloud bursting" strategy for maximizing ROI: Invest in a private cloud and supplement with public cloud services when utilization rates reach a certain point. With more than 50 million compute hours, 100 TB of data, and 500 million files in a typical DreamWorks movie, doing all of it in Amazon's cloud would be too pricey. Instead, only 20%--10 million CPU hours--of Kung Fu Panda 2 was hosted in public cloud facilities. DreamWorks built most of the capacity it needed and farmed out only what it couldn't handle--the burst. Zynga is taking a similar hybrid approach; on a recent earnings call, COO John Schappert said that 80% of the gaming company's daily compute usage was in its private cloud, with the balance on Amazon.
The nature of a hybrid makes ROI calculations that much more complicated. But you can nail down a number, as we'll discuss. The reality is that most enterprise organizations will use the hybrid model for several reasons, including the aforementioned cloud bursting, and the innate complexity of a nongreenfield enterprise. These organizations have a much harder time changing their computing model compared with startups.There's an IT joke that goes, "The Lord could create the world in six days because he didn't have an existing user base." Startups can afford to be "do or die" in the cloud, so enterprise IT teams shouldn't beat themselves up unduly if they see startups running rings around them in using cloud services. Instead, as we discuss in our full report, established companies should focus on the middle ground, where you build capacity for what you know you're going to use and plug in on-demand for the portion that can't be forecast.
Still, no matter how you cut it, the new IT ROI calculation includes cloud. Case in point: Joe Emison, VP of research and development for BuildFax, a provider of U.S. property history data, is responsible for an application that serves 70 million records to customers across the United States. "When things fail, the people who are administering the setup are developers, not infrastructure guys," Emison says. "If you can redeploy it anywhere, hardware failure doesn't matter." If a server goes away, the system is architected so that he just logs in from wherever he may be, hits "relaunch," and he's back in business.
Download the March 2012 InformationWeek digital supplement