How data centers cope with the economic downturn

With the economic downturn, data centers face serious budget cuts. Recent survey data indicates they have turned to server virtualization to reduce server spending and power consumption as well as IT automation to address staff reductions.

Between February and March 2009, conducted a survey on how data center managers are coping in the economic downturn. Subscribers were contacted by email and 235 end users, primarily in U.S.-based IT shops, responded.

For more on data center budgets, staffing and more:
The 2008 data center purchasing intentions report

How willl data centers weather the economic downturn?

Observations on the IT job market: The government sector
Special IT projects survey

Survey finds data centers understaffed

The data shows that IT budgets took a major hit in the first quarter of 2009. Nearly 70% of the respondents said they face budget cuts, and one-third must deal with budget cuts of more than 20%.

The largest cuts across IT budgets come from staff reductions, followed distantly by reductions in spending on server hardware, application software and systems management tools, respectively.

Despite the cuts, demand for IT and data center services keeps growing. And about 30% of respondents are proceeding with all planned data center projects. Twenty-five percent have canceled some data center projects, 17% are scaling all projects back, and only 4% are canceling projects altogether.

Crafting a survival strategy
In the face of increasing compute demand, how are cash-strapped IT organizations getting by? Primarily by making their servers last longer in production. Servers are typically replaced every three years. Two-thirds of IT shops have extended the production life of server deployments in 2009. More than 35% say they'll keep servers in production for six months to a year longer, 34% say they'll extend server lifecycles by two years.

IT organizations are coping with the downturn by making their servers last longer in production.

IT pros said prolonging server work life is a viable option. "This is definitely a workable strategy; systems don't suddenly stop working when they're depreciated off the books. The three-year-lifecycle is more of an accounting issue than a matter of hardware reliability," said Bill Bradford, the senior systems administrator at an energy services firm in Texas.

"As long as a server is doing what it is supposed to do and is keeping up with workload, I don't see anything wrong with keeping it around as long as warranty service and/or spare parts are available. It's an extreme case, but I've seen 12-year-old systems in production simply because they still worked just fine, plenty of spare parts were on hand, and there was no driving reason to put their functionality on a newer platform."

Kyle Rankin, senior systems architect for QuinStreet, Inc., a Foster City, Calif.-based marketing firm, said the strategy is especially applicable if a company plans to cut back on all IT operations.

"If companies have scaled back, it's altogether possible the current servers are handling less load, and you might be fine with them for some time. The main thing to worry about after three years is renewing support contracts. Some businesses assume and pay for the standard three years of support and then afterward they assume they will replace the gear. In this case, they'll hold onto it so they'll have to decide whether they will make extra payments to keep the hardware supported, or risk having it fail when it's older and more likely to fail."

"Another thing a lot of people seem to be doing is looking for free [that is, open source] alternatives to software they currently have," Rankin Said. "I've seen that happen quite a bit. When the time comes around to pay software licenses for something for another year for a proprietary product, people start looking for alternatives."

Still some experts say while this strategy can stave off capital expenditures now, it can be costly in the long run.

"Unless you have very sophisticated mean time between failure predictions, or spend a lot of time, effort and some money on preventative maintenance (fan replacement, power supply replacement, etc.), operators could find themselves fighting fire after fire as their servers go beyond their useful life and start an increasingly frequent series of cascading failures," said Andi Mann, the vice president of research for Enterprise Management Associates.

Joe Clabby, the president of Clabby Analytics, said this is fairly standard recession behavior. "I've seen this happen before in economic downturns. Generally it means that companies go into stasis. They maintain life support for the existing suite of applications and databases that they run and don't add a lot of new workloads. Sometimes they push utilization rates up on their servers to accommodate new workloads, moving from 5% utilization to 10%."

Better server use is in fact, the second piece of the puzzle. More than 35% of survey respondents say they're buying larger servers and consolidating virtualized workloads onto higher-end platforms.

"Two parts of the server architecture that take the biggest hit when companies are being very aggressive about consolidating workloads are memory and bandwidth," said Charles King of Pund-IT Inc. "Higher-end servers have more memory and back-end bandwidth options than one- and two-way servers. That seems to be the platform of choice when you're trying to squeeze a number of workloads onto a single box."

Server virtualization spending continues, but VMware put on notice
In fact, server virtualization was the one area where people were hesitant to make cuts. Only 15% said they'd spend less on server virtualization in 2009. Around 38% of respondents said their spending on server virtualization was unchanged, 28% said they would increase spending, and 19% do not use server virtualization.

Aaron Sawchuk, the CTO of Rockland, Mass.-based ColoSpace Inc. recently told said that there is a much heavier focus on virtualization than on server hardware in his facilities. "In general terms, we continue to see customers making purchases that can be directly attributed to cost savings, [especially] within the virtualization space, where they can reduce the overall data center footprint," Sawchuk said.

In's August 2008 data center purchasing intentions report, VMware's lead in server virtualization seemed unshakable, with a dominant 64% of respondents using a version of its software. VMware ESX 3.5 alone had 37% market of share, and the next non-VMware deployment was Microsoft Virtual Server, at 7% adoption.

Forty percent of respondents use VMware virtualization but are considering alternatives because of price.

But in this economy, that landscape has changed significantly. Already half of respondents said they use an alternative to VMware in some capacity, and another 40% said, they use VMware but are considering alternatives because of price.

Of the VMware alternatives, 44% of respondents found Microsoft's Hyper-V or Virtual Server most appealing, followed by 26% for Citrix Systems' XenServer, 15% for Red Hat KVM, and the remaining 15% went to other platforms.

According to Mann, server virtualization is a commodity, so the temptation to dump a costly VMware solution for a very cheap (or free) Hyper-V or Xen-based solution is quite logical. But it can be risky.

"The value from virtualization is not the hypervisor, but how well you can manage entire virtual systems. Better virtual systems management directly contributes to faster mean time to recovery, higher VM [virtual machine] density, lower admin staff costs, and greater physical resource utilization," Mann said.

The problem, Mann said, is the alternative virtualization platforms have limited management tools available. So while the up-front costs for both Xen and Hyper-V will be much lower, companies should be aware of longer-term total-cost-of-ownership issues.

Clabby disagrees. "Hyper-V is better integrated into a Windows strategy than VMware, because Hyper-V tucks in under Microsoft's management products and can accommodate virtual and physical management -- as opposed to just virtual management," Clabby said. "Hyper-V will overtake VMware over time. VMware is not widely adopted worldwide -- and Windows is. VMware is expensive -- and then they start driving you to buy more and more of their infrastructure and management products."

And buying more management products isn't something IT organizations are eager to do.

IT management spending, CMDB on the skids
More than half of survey respondents said are less likely to spend on systems management tools in this economy. So IT management software implementations would need to have fast and significant return on investment to get a green light.

When asked where they would spend their meager IT systems management budgets, respondents said 35% would go to monitoring software. But in a surprising reversal, historically low priority IT automation tools took the No. 2 position, with 20% of respondents saying they would prioritize software that could help them automate IT processes and cope with less staff.

[Configuration management databases] are viewed as nice to have.
Joe Clabby,
presidentClabby Analytics

Configuration management databases (CMDBs) tools took a backseat. In the 2008 survey, nearly 80% of respondents either had or were evaluating a CMDB, and 35% of respondents said they would increase spending in 2009 on a CMDB project. But according to the latest data, that momentum has stalled.

When asked which systems management tool they would be least likely to spend against, 21% of the respondents said CMDB software, followed by capacity planning tools with 18%.

"CMDBs are viewed as nice to have," Clabby said. "CMDBs greatly simplify management and troubleshooting, but most IT managers focus on the management of physical devices -- so they already have a lot of the software that they need."

According to our 2008 survey data and Mann's organization's research, CMDB deployments can often show positive ROI quickly. But Mann says CMDB implementations can require two to four full-time employees, plus software costs, which might not be available at the moment.

"But with significant benefits and payback, it would be hard to drop a CMDB system project that was already under way and starting to show results," Mann said. "In some cases, we are seeing budgets shift -- funding CMDB as OPEX instead of CAPEX. But we are not really seeing enterprises abandon current CMDB system projects overall."

The price of power catches up with IT
Electricity costs have finally caught up with IT departments, as 80% of respondents said they are more likely to pursue energy-efficient IT in the down economy. That's good news for the environment and corporate pocketbooks.

Of the handful of respondents who are less likely to pursue energy-efficiency initiatives, they cited additional capital expenditures as the main barrier to the projects. But 64% of the respondents said they see green IT as a cost saver, not a cost center.

Server consolidation and purchasing more energy-efficient servers are the most popular means for data center pros to improve efficiency, followed distantly by cooling system improvements, auditing and unplugging unused servers, and implementing server sleep or shutdown tools when servers are idle.

"I think if people didn't take efficiency seriously before, they will now," said Christian Belady, a principal power and cooling architect for Microsoft's data centers. "When profits are high, it may be hard for some organizations to justify spending much time on driving efficiency. In environments with little or no profits, the efficiency of your operations could make the difference between a profitable year and a loss."

Matt Stansberry is the executive editor of

What did you think of this feature? Write to's Matt Stansberry about your data center concerns at [email protected].

Dig Deeper on Data center budget and culture