At a Gartner conference in November 2007, more than one-third of the attendees said their newest data centers are seven years old or older, meaning they weren't designed for the power and cooling needs of today's high-density servers. Half of the respondents predicted that they would need to expand their data centers over the next three years, with many saying that they expected to go "from one to two data centers, in part to address disaster recovery," according to Gartner.

Whether they're building new data centers from the ground up or revamping existing space, data center managers are virtualizing multiple applications onto single physical servers, consolidating many data centers into one or two facilities to save on real estate and other infrastructure costs, and redesigning the physical arrangement of servers to save on cooling costs.

"We're doing a thermal heat analysis to understand where we have hot spots, and implementing hot and cold aisles" between racks to prevent cool air from being contaminated by exhaust heat, says Jim Lowder, vice president of technology at health care provider OhioHealth in Columbus.

Since starting an overhaul of its data center in 2003, the Pennsylvania attorney general's office has achieved close to "lights-out" operation of 130 servers, more than 1,000 computers and 20 offices across the state. It uses virtualized servers and remote management software to slash the time and effort required for routine functions. "We can completely rebuild a server and have it ready to roll out in about seven minutes," says data center administrator John Nester. He has also kept head count and energy use flat since 2005 while more than doubling the number of servers.

For his part, Richard Balentine, now the CIO at Greensboro, US-based NewBridge Bank, says he had to interrupt a planned upgrade of his application infrastructure at Lexington State Bank to cope with the 2007 merger between Lexington State and FNB Southeast that created NewBridge Bank.

The original modernization effort included upgrading applications such as customer relationship management systems and bringing outsourced software back in-house. Among other upgrades, Lexington State installed its first storage-area network, an EMC Corp. system, and upgraded its network.

But in early 2007, the looming merger required Balentine to refocus on consolidating the two banks' infrastructures, which included reconfiguring FNB's data center in Reidsville, as a disaster recovery site. Consolidating and combining bank operations took seven months and included expanding the SAN and deploying new, virtualized servers using software from VMware.

Creating -- or reducing the cost of -- a disaster recovery site is often part of a data center upgrade plan. For example, once it has moved to its new facility, Christus Health will move its disaster recovery site from an outsourcer and into its new data center.

Some vendors, as well as customers, argue that data center managers shouldn't just look for cost savings as they reconfigure their centers; they should aim to fundamentally transform how they deliver IT services.

The Pennsylvania attorney general's office has virtualized servers with VMware software, moved from direct-attached storage to network-attached storage from NetApp Inc. and upgraded its network to 10 Gigabit Ethernet. Power capacity and cooling weren't issues, since the data center was built to handle mainframe needs and is, if anything, too cold rather than too warm. Nester says he is using two-thirds less space than he did before the upgrade.