FRAMINGHAM – Despite the consolidation trend, some companies are continuing to branch out with their data centers for reasons including redundancy of the centralized data center for backup, a more comprehensive disaster recovery plan and better uptime at regional offices.
Indeed, some companies actually prefer this approach, and experts say it makes sense for organizations such as those with branch employees who can act as IT technicians or firms with locations dispersed on different continents.
“The satellite data center model is alive and well in companies that support multiple regional offices, such as banks, chain stores and insurance companies,” said Charles King, a principal analyst at Pund-IT Inc. in Hayward, Calif. “The benefits are substantial for any industry where it’s important and advisable to keep workers close to data and their IT infrastructure.”
George Hamilton, an analyst at Yankee Group Research Inc., said he’s also in favor of satellite data centers in some instances.
“The advantage is leveraging existing space, rather than having to build additional data center capacity,” he said. “With virtualization, you can make the satellite data centers appear as one big data center — the users have no idea where the physical devices reside, nor do they care. It also builds in local redundancy — unless the entire building is affected. You can also shut down one satellite data center and have another pick up the functionality during technology refreshes and upgrades.”
Three kinds of satellites
One type of satellite data center is a “data room” on each floor. This was formerly known as an intermediate distribution frame (IDF) and is now commonly called a tenant technology room (TTR). Granted, these are not true data centers. However, the TTR of a large organization might be larger than an entire data center in a small company.
A second type is a regional office with a local data center that provides a subset of services, such as e-mail and document servers. Law firms are well known for having data centers at each office, serving just a few hundred employees.
A third is a data center located in a different state from the main data center entirely, usually to reduce constantly escalating utility costs. In this approach, the IT organization might be relocated to the remote data center, with the home office staying in the original location.
Business continuity is key
In all cases, decentralizing can help organizations minimize the consequences of a catastrophe, said Steve Novak, CIO at Kirkland & Ellis LLP in Chicago.
For some mission-critical applications, such as e-mail and instant messaging, the firm runs a data center in each office that is redundant with the centralized data center. “If we lose e-mail in the local office, we know the secondary server is available and vice versa,” Novak said.
In the Kirkland & Ellis model, a 100-square-foot TTR on each office floor handles basic networking, a 1,000-square-foot satellite data center handles mission-critical services such as e-mail, and a 10,000-square-foot global data center in Chicago handles most companywide services. T
hat said, however, Kirkland & Ellis stores all documentation — the intellectual capital that forms the lifeblood of the company — in a centralized data center where long-term redundancy is more important than uptime. The other option, which is less reliable, would be to store documentation in each regional office.
With a move to server virtualization, the firm is building a global data center next year that will further unify data services but still not change the regional data center strategy with regard to e-mail and instant messaging.
Power savings too
Another reason some companies use satellite data centers is the escalating cost of power, especially in New York and California.
Ken Brill, founder and executive director of the Uptime Institute in Santa Fe, sees a trend with data center being located in a different state from headquarters, or even in another country such as India.
The Google Inc. and Microsoft Corp. model is to build server farms in states where energy costs are low, Brill explained. These sites do not always provide data services for the corporation itself — in Google and Microsoft’s cases, the server farms are intended to deliver improved performance for external customers. But Brill noted that even major data centers don’t necessarily have to be located at a company’s headquarters.
One Uptime Institute client that Brill declined to name is building a global data center in a market where utility costs run only about 4 cents per kilowatt-hour. Compared with New York, for instance, which costs around 15 cents per kWh, that’s a savings of US$30 million per year.
In general, Brill advises customers weigh the risks of data center location carefully. Centralization reduces management and staffing costs, yet “the more centralized you are, the greater the consequences from a single event,” he said. “The challenge for CIOs is to weigh the economies of scale with reliability.”
Brian Babineau, a senior analyst at Enterprise Strategy Group in Milford, Mass., said that the main reason companies build regional data centers is to save power.
Further, Babineau said, when a company builds a new satellite data center, the company typically uses it only for data, not for telecommunications. The customer leaves its existing equipment at the main data center until it has moved to IP. When that happens, the satellite data center can handle both voice and data, all in one space, making it more efficient.
An added benefit, according to Babineau, is that because the computing load can be spread among multiple locations, power bills can be reduced in each location, particularly if the satellites are in areas with lower power costs.
Babineau said the biggest downside to regional data centers is that customers have to disperse the workforce. But the cost savings from locating a data center in a relatively low-cost area such as Arizona or the Pacific Northwest could outweigh the detriments of having employees spread out across the country.
Of course, there are higher management and staffing requirements in a decentralized environment — reasons that one Gartner analyst is vehemently against the notion. “Gartner’s view is that clients need to move to the smallest number of data centers and make them as large as possible — and to limit the number of satellite locations,” said Rakesh L. Kumar, an analyst Surrey, England. “This way, an organization is able to streamline its people resources and operational processes.”
Steve Sams, an IBM vice president for site and facilities services in New York, agreed. “We do not see a trend in moving to satellite data centers,” he said. “In fact, the trend is just the opposite, with customers moving toward data center consolidation.”
Sams pointed to an example with an unnamed client company that went from 32 data centers down to just two and is now saving $180 million annually in operational costs. He also referred to a Gartner study that indicates that 57% of those surveyed plan to consolidate in the next 12 to 18 months.
“In most cases, it is easier to define and manage a set of best-of-breed operational procedures across a small number of data centers versus a large number of data centers,” Sams added.
Kumar wondered about the $180 million savings for th