Data center consolidation is all the rage — enterprises are taking geographically distributed data centers and collapsing them into one or two centrally managed locations that are less expensive and easier to operate.

But data center consolidation isn’t for everyone. Analysts say there are numerous reasons to maintain a distributed model. Bucking the trend to combine locations, some companies are building new data centers or adding to their existing footprint by renting more space in co-location facilities.

Disaster recovery is one of the most common reasons that enterprise stick with multiple data centers, says analyst Arun Taneja of the Taneja Group.

“When you have multiple data centers they can act as disaster recovery sites for each other, and therein lies the value,” Taneja says. “Having one data center is never going to be enough. I’m going to have to have a disaster-recovery site whether it’s my own, or I borrow from somebody.”

Real estate prices sometimes make building another data center the right option. Say a company in a large metropolitan area such as Manhattan has run out of data center space and capacity for power and cooling. Expanding into another floor in a Manhattan high-rise is prohibitively expensive, so it’s considerably less expensive to build an extra data center in New Jersey, says Tom Atwood, systems group marketing manager for Sun Microsystems. 

Concern about latency is another reason for sticking with geographically distributed data centers. Vendors such as Citrix say they enable data center consolidation with speedy application delivery systems, but in general it’s better when data is used near where it is created, Taneja says.

Even companies that are consolidating should consider whether they have systems that must remain distributed, says Forrester analyst James Staten. “You may have certain services or applications that require rich interfaces with a client, which is not all that easy to do over a wire from a headquarters data center,” he says. “In that case, it makes sense to have geographically dispersed data centers.”

Banks and telecom companies have extremely rigid latency requirements that usually necessitate a distributed model, Staten says. Global financial services institutions may also need data to stay in specific geographies to comply with government regulations, he adds.