New technologies mean shorter server life cycles

Are you up to speed?

1 2 3 Page 3
Page 3 of 3

Beyond virtualization

Although virtualization is the biggest force accelerating server replenishment, it's not the only trend shortening life cycles. A need to meet the demands imposed by new and more powerful software -- particularly applications incorporating multimedia, analytics or other processor- and memory-intensive operations -- provides another strong motivation for upgrading servers more rapidly.

"It may be that the latest version of an ERP system delivers so much functionality and business efficiencies that moving up a purchase seems to be the right thing to do," Stahl says. "Maybe the business is struggling with lowering their costs and improving their margins" and making this decision is a good option.

For his part, Nowak feels that upgrading servers in response to the availability of new or enhanced software is something of a no-brainer, since hardware cost usually isn't a critical factor. "Hardware is not the major expense from an IT standpoint -- it's less than 15% of our budget," Nowak says.

servers

"Most of our budget is [software] licensing and consulting services, so if there's a new release of Microsoft, we're going to go with it and upgrade the hardware while we're at it. It's a lot easier to move from one server to another rather than try to do an upgrade in place."

A seemingly endless array of new or upgraded regulatory mandates, typically arriving in waves, is another factor behind faster server turnarounds. "Government compliance and those sorts of issues -- there's no way to track these mandates other than keeping your ear to the ground until something becomes law, and it usually requires you to move pretty fast," Stahl says. "HIPAA, SOX and the others, they all have an impact from a technology-buying standpoint."

The accelerating pace of hardware evolution is also inspiring many managers to chuck inefficient older servers in favor of more powerful replacements. Bourassa of EmpireCLS cites his HP blades as evidence of how advancing technology is inspiring many data centers to speed up their server refreshment cycles.

"When we bought our first wave [of servers] almost three years ago, they only supported 32GB of memory," he explains. "Now, the latest G6 models have 96GB of memory, so I can get triple the capacity for my virtualization."

Settling in

A three-year replacement cycle has become the norm at most data centers, says Paul Prince, CTO of Dell's enterprise product group in Round Rock, Texas. "Three years is kind of a nice sweet spot," he notes.

Staying on top of IT industry developments is key to spotting and leveraging important trends, Bourassa says. "Make sure that you are attending your hardware vendors' conferences, meetings and road maps so that you can make very good decisions around when to make your procurements."

Questioning vendors on their upcoming product offerings is crucial to ensuring that systems acquired at a cycle's start incorporate the latest features, Bourassa says. "If [a needed feature] is a few months out, you can postpone your procurement to the next generation." Bourassa says that it's important to have a vendor you can trust.

That said, he expresses some disappointment in his vendor's willingness to disclose critical information. "HP did keep things a little close to their vest, [such as] the large capacity change they were doing in their server blade farm," he says. While the old models supported 32GB of memory, the new servers accommodated up to 96GB -- up to triple the amount. "So I did get a little caught off-guard when they went from their [ProLiant] G1 model to their G6 model," Bourassa says.

He says he wishes HP had alerted him to this pending model change sooner. "I probably would have slowed down my G1 procurement to wait for the G6s," Bourassa adds.

HP had no comment for this story.

Prince, meanwhile, observes that every enterprise is different and that the only reliable way of determining what length and type of server refreshment cycle works for a particular data center is to do some serious number-crunching. "You'll have to run your own numbers to determine what's best for you."

John Edwards is a technology writer in the Phoenix area. He can be reached at jedwards@gojohnedwards.com.

Copyright © 2010 IDG Communications, Inc.

1 2 3 Page 3
Page 3 of 3
  
Shop Tech Products at Amazon