LAS VEGAS -- With the recession under their belts, some data center managers have moved toward longer server refresh cycles. Instead of the three- to four-year cycles that were common before 2008, they have pushed the interval to five and even six years.
When it comes to server refresh cycles, “there’s no absolute law of physics that impact the number of years you picked,” said Carl Claunch, the vice president and distinguished analyst for Gartner Research at the Gartner Data Center Conference. When the recession hit, many enterprises decided to postpone server refreshes for a year or two, “and we all lived another year, and the world didn’t go into flames.”
The widespread deployment of virtualization has further emboldened data center managers in this direction. “We’re not buying as much white space as we were,” Claunch said, referring to virtualization’s ability to boost server utilization.
IT director of a large retail food chain
Another Gartner analyst went one step further: to consider upgrades in place – upgrading servers with the latest high-powered multicore chips, rather than doing a wholesale server refresh. Processor manufacturers keep releasing ever more cores per processor, driving compute densities, said David Cappuccio, Gartner’s managing vice president and chief of research for infrastructure. “Perhaps there's no money to build, but is there money to retrofit?” he said.
Virtualization protract server refresh cycle
Conference attendees, however, were mixed on these suggestions. On the topic of server refresh cycles, one attendee, an IT director at a large retail food chain, said he agreed server refresh cycles had lengthened by about a year, thanks to virtualization.
“VMware is our refresh platform,” he said. As demand for new applications comes, new blades are brought intothe virtualization resource pools, and old blades fall off the end.
But for others, existing business processes would make it difficult to consider longer refresh cycles, much less in-place upgrades.
Simply put, “We lease all our equipment,” said the IT director at a Fortune 100 manufacturing firm, eliminating any flexibility surrounding server refresh cycles and in-place upgrades.
Likewise, accounting processes could thwart changes such as in-place upgrades, said John Toner Jr., the associate director for IT at Verizon Wireless data center site operations. “We wouldn’t do that because of capital depreciation rules,” Toner said, explaining that it would extend a server’s life in the depreciation cycle. Further, the data center team is wary of overly dense virtualized servers, lest they take up less floor space, he said.
“If I virtualize, I start to lose floor space, and if I lose floor space, it prevents me from putting something else in.”
But some smaller organizations save money by upgrading server processors rather than refreshing the entire platform. Matt Lavallee, a director of technology at MLS Property Information Network in Shrewsbury, Mass., told SearchDataCenter.com in October that his firm upgrades its Hewlett-Packard DL585s with new AMD processors once a year. “We do this in particular with our eight-way boxes, where there is so much investment in the total asset and [where] the upgrade is one-tenth the cost of replacing the server for the new speed,” Lavallee said.
Dig deeper on x86 commodity rackmount servers