The whopper data center bets they must make still pivot around brick and mortar. Here are a few companies that made very different decisions to meet their data center needs in the past year--insource, sell, and build.
(For additional insight into data center strategy, see also Bob Evans' Global CIO column, "Data Centers Behaving Boldly: Meet Tech's New Rock Stars")
Insource A Data Center
Whitney National Bank decided to insource a data center, under the leadership of Scott Erlichman, senior VP of technology infrastructure for the regional bank. The bank's 3-year lease was up on co-location space it used for disaster recovery, and the bank found rates had risen 50% or more since 2006, because demand for such space is high. At the same time, the bank already was planning some construction at a site it owned in Alabama, in order to do some back-end work such as check processing. That made insourcing an intriguing option.
The project turned out to be more complex and costly than initially expected, says Erlichman. The building itself, and the heating and cooling infrastructure in particular, weren't up to the standard needed for a modern data center, so that added to the projected cost estimates. That pushed out the expected return-on-investment, but the numbers still added up to go ahead.
Since October, Whitney Bank has been running the Alabama data center as a lights-out facility, with the computing infrastructure managed entirely remotely, mostly by staff in New Orleans with some oversight from its Dallas and Houston offices. There's someone on-site who monitors heating and power as needed, but that was required for the back-office processing, so didn't mean additional staff. No IT people work there, though there's staff within a few hours drive if needed.
Whitney Bank didn't treat the insourcing move as a blank technology slate; in fact, it basically fork-lifted the IT infrastructure, including the storage used to support specific apps at the outsource facility, and moved it Alabama. Since that involves confidential banking information, it was no small matter--think secure transportation and police escort. Over the next two years, however, the bank will do some application redesign that will let IT bring in more new technology. The bank's also looking into new application monitoring technology, so IT has more insight into performance the business units value--say, the time it takes to complete a transaction, rather than the availability of a server or network. It's already largely tapped out the gains to be had from virtualizing servers, but it's looking at increased use of blade servers for some further efficiency gains in management.
Even as it insourced, Whitney kept its core data center with an outsourcer, AT&T. That's a high-performance data center, and the bank didn't feel it had the operations expertise to run that. However, it did use virtualization to reduce its footprint at that facility by 40%, which cut its costs.
As companies strain against data center capacity, they face a choice: build, retrofit, or outsource, says SunGard Availability Services data center consultant Mickey Zandi, who worked on the Whitney bank project. Zandi says he sees more companies looking at "smart retrofit" projects like this one, with an in-house data center paired with an outsourced one, with each backing up the other. They want "flexible infrastructure."
Three big things Whitney did right, says Zandi. First, it had an integrated project management office for all functions under the CIO's office--covering elements including construction management, data center design and relocation, commissioning and testing, and planning of heating and cooling systems. Second, it had top executive oversight and support, including Erlichman's. Third, the IT team understood its infrastructure, even though it had been outsourced, so they were clear on the interdependencies of apps, systems, and networks. "If something happens, then one needs to know exactly who is going to be impacted, what line of business, what end user client, and what is going to be the impact on the business," Zandi says. Most companies, Zandi says, don't understand those interdependencies. Sell A Data Center
One Fortune 500 company decided to take a somewhat unconventional approach to outsourcing its data center--by selling its New Jersey data center lock, stock, and barrel to the India-based outsourcer HCL for $15 million. The company, which HCL didn't name in announcing the move, contracted with HCL for infrastructure services. HCL, meanwhile, will use the data center not only to serve the company, but also to meet the growing demand HCL has for U.S.-based, outsourced IT data center services. "Our clients are looking for data centers where they are," says Shami Khorana, president of HCL America.
HCL does most of the remote monitoring needed for the data center from India. But companies don't worry much where the people working on the data center are, Khorana says. In financial services and pharmaceutical industries, in particular, it's the data they want to remain in the U.S. "Even for cloud computing, it's very important to clients where that data is residing," says Khorana.
These kind of deals may appeal to CIOs trying to cut as much IT capital expense as possible from the budget. However, a lot of stars need to align to make such a deal happen. Outsourcing that merely turn company employees into outsourced employees have a spotty record. The difference in this deal is HCL's using the data center as a base to support other clients, spreading fixed costs over multiple clients. For a CIO, finding an outsourcer that needs its data center capacity, and also is the right strategic partner to work with, is a nice match but tough to find.
Build A Data Center
For the most sophisticated users of computing power, building brand new data centers is still in style.
The CME Group, which runs some of the world's largest commodity exchanges, not only decided to build, it went big. CME runs the Chicago Mercantile Exchange, Chicago Board of Trade, and the NYMEX, in a world where trading clients measure success and failure in milliseconds. It has built a 260,000 square foot data center in the Chicago suburbs, while keeping the data center capacity it has in downtown Chicago, and New York.
Hewlett-Packard in late 2008 finished building six brand-new data centers in the U.S., part of a global consolidation effort to eliminate 70-plus smaller data centers around the world. CIO Randy Mott's strategy is to have the heavy computing done in hyper-efficient, large-scale data centers, and have employee's PCs around the world access them.
And of course, those that aspire to grand cloud computing ambitions, the likes of Google and Microsoft, have built data centers on a massive scale. Facebook is the latest to join the crowd, breaking ground on an Oregon facility.
My point in sharing these examples is to show the range of strategies CIOs are exploring in their data centers, and how there's no one simple option, from build to rent to sell to share to whatever new approaches you're cooking up.