Cloud computing emerged as disruptive technologies often do—in an entrepreneurial spirit of collaboration in which resources and costs are shared. But so is data, inadvertently or not. When Citigroup embraced the cloud, the cost-saving benefits were obvious, but so was the risk to the firm's enterprise. There was never any question about taking its cloud private.
"Facebook has more data than we do," says Antonio "Yobie" Benjamin, chief technology officer at Citigroup's global transaction services in San Francisco. "Citi has roughly 250 million accounts. Facebook has about 900 million accounts and trillions of photos. A complaint Facebook is likely to get will be along the lines of, 'Dude, where's my photos?' With us, it's 'Where's my money?'" For Benjamin, that's an essential difference.
Building a private cloud, however, is no small endeavor for an organization of Citi's size, and that's one reason the firm is recognized by U.S. News' as one of America's Most Connected Companies. Citigroup operates in 100 countries, has $13 trillion in assets under management, and moves $3 trillion worth of transactions every day. "That's roughly a quadrillion a year," says Benjamin. "There's nothing here that can be left to chance." When it comes to doing business with other cloud-based firms, he says, "If you have a public cloud, don't talk to us."
Citi began its move to the cloud in 2010, beginning with a handful of employees who started work on an internal infrastructure that evolved into the private cloud model. This in-house approach may be costly, but the security benefits are obvious: A public cloud's services are available to anyone who can pay and where the risk of commingling data is high. A private cloud is proprietary; the owner decides who uses it and how, as well as the level of security.
Still, Benjamin says cloud computing is essential for Citigroup. "The benefits that Citi gets are, in a word, elasticity," says Benjamin. "When you need computing power, it's there, so you gin it up. When you don't need it, you gin it down. You can adjust that along the information-supply chain, depending on the processing speed or the bandwidth that you need." The result is economies of scale—and money saved.
When he's not worrying about security at Citi, Benjamin worries about regulators whose demands are significant, especially concerning data retention. "We can't discard any of it," he says, "because we have to be prepared for the day when a regulator waltzes in and says, 'Can I see the last 10 years?' And we say, 'The last 10 years of what?' That's when the fun begins."
Cloud data—the type a government regulator might request—allows data to be retrieved and analyzed more quickly and in more flexible formats, although mainframes could still be used as a backstop for secure data storage.
[In Pictures: The 10 Youngest Billionaire's in the World.]
Security is a constant worry. "Every day, we repel 40,000 attacks against our system," Benjamin says. "You never hear about the ones we keep out, only the one that gets through."
Speed matters as well. High-frequency trading, which requires sending trade data to stock exchanges, will be the model for data speeds at Citi and elsewhere. "We want to go past the millisecond level to the microsecond level," Benjamin says. "Milliseconds are the speed at which your credit card is approved. So when your card is swiped at Starbucks in 500 milliseconds, which is basically half a second, you get your latte. That's not fast enough for high-frequency trading or any high-performance computing. That's immensely slow. The goal is to move to nanoseconds. So that credit card swipe would speed up 100-fold."
Citi's high-frequency trading initiative supports algorithmic trading in equities, foreign exchange, and asset classes in which large trades are often chopped up into smaller ones for quick execution. It also provides services to hedge funds.