Greening the Data Center
Energy efficiency isn't just about using windmills and solar
panels. CIOs can reduce their power costs — and help the
environment — by adapting virtualization and other innovative
processes and technologies.
By Rusty Weston
CIOs are going green. Around the world, IT departments are embracing new, energy-efficient technologies and practices. Despite concerns about underutilized servers in distributed computing environments, much of today’s “green IT” transformation is happening in the data center. For good reason: Data centers are a significant node on the energy grid. In the U.S. alone, data centers burned through 60 billion kilowatt-hours (kWh) in 2006, or roughly 1.5 percent of the country’s total electrical consumption for that year, according to the U.S. Environmental Protection Agency.
Looking ahead, the agency predicts that energy consumption by U.S. servers and data centers will nearly double in the next five years, reaching more than 100 billion kWh and costing businesses roughly $7.4 billion a year.
While computing hardware consumes a great deal of this power, a surprising amount goes to keeping the data center facility up and running. “For every dollar you spend on your data center, another dollar is wasted just keeping the thing cool,” says Andy Lawrence, Research Director for Eco-Efficient IT at The 451 Group, a market research and analysis firm.
Trouble is, much of this power consumption flies under the radar screens of CIOs. That’s because electric power bills typically go not to IT, but to facilities management. According to a study by the Economist Intelligence Unit, just 42 percent of IT executives say their organization monitors its IT-related energy spending. As the saying goes, you can’t manage what you can’t measure. So IT is neither held accountable nor given management incentives to hold the line or reduce energy consumption.
Yet awareness among CIOs is growing, thanks to skyrocketing energy prices, growing concerns over global warming, a series of worldwide droughts and the general prominence — on both the political and cultural fronts — of the need for greater energy conservation. Many CIOs also have a floor-space problem: While they’re being asked to do more, they are not being given additional funding for extra data-center space. “The top priority is data-center efficiency,” Sabet Elias, CTO of investment bank Lehman Brothers, told
InformationWeek earlier this year. In 2006 Lehman Brothers boosted its energy efficiency by 25 percent; it has set its goal for 2008 at an additional 35 percent boost.
That’s especially true in Australia, where a continent-wide drought began in 2002 and shows no sign of easing. The massive lack of rain is hurting farmers, causing food prices to rise and generally raising the country’s awareness of the need to conserve both energy and natural resources. Also, because an estimated 75 percent of Australia’s electricity is generated by burning brown coal — considered an environmentally dirty source — the country’s CIOs are under pressure to reduce their electric power needs.
Smaller Footprint
Some already are. In 2007 one of Australia ’s largest banks publicly stated that it would reduce its environmental “footprint” — water, waste, electricity and greenhouse emissions — by 5 percent before year’s end. As part of that effort, the bank’s IT department banned the use of screensavers on the company’s 31,000 PCs, saying the screensavers prevent PCs from switching off when unused. This screensaver ban is expected to save at least the equivalent of approximately $430,000 a year, or about 4 percent of the bank’s total energy use. The Australian bank also plans to install centralized management software that will automatically switch nighttime PCs into hibernation mode, lowering energy costs by another 4 percent. In addition, the bank is switching all systems from CRT monitors to LCD flat screens, which consume roughly half the electric power of the older units. Finally, the bank is looking into virtualization software as a way to reduce the number of servers it runs—and to reduce its data center’s electric power needs.
One surprising development is the emergence of the mainframe as a green computing device. Unlike cars — where a smaller footprint coincides with better fuel economy — big computers are actually more energy-efficient than midrange and PC-based servers. “It’s all about embracing energy efficiency for cost savings,” says David Anderson, a self-described green evangelist at IBM. “It comes as a shocker, the mainframe being energy-efficient, but we didn’t get there in a day.”
That leads industry experts like William DiBella, president of AFCOM, an organization of data center professionals, to predict that mainframe sales will “definitely increase.” But Rakesh Kumar, a VP at analyst firm Gartner, isn’t so sure. “The mainframe’s strength is that it is an energy efficient, high-transaction, highly available workhorse,” he says. “But if you want to move different apps to a legacy platform, you do so because it’s an appropriate platform. Energy is just one consideration.”
But while the mainframe generally Draws most of the attention in the data center, midrange systems also contribute to a green strategy. According to Mark Monroe, Director of Sustainable Computing at Sun Microsystems, one problem with most data centers is they tend to overprovision cooling systems—burning excess watts in systems that should be consolidated or run virtually. To show how it can be done more efficiently, Sun is showcasing three data centers — in Santa Clara, Calif.; London , England; and Bangalore, India— that use virtualization techniques to drive server consolidation — one key way to reduce energy consumption.
The gains from consolidation can be considerable. By 2010, the cost to power a server over its lifetime will equal the cost of purchasing that server, predicts IBM. “It’s a double hit,” says Anderson, “because you also have to have power for the infrastructure, the air conditioning, storage and other systems.” That’s the reasoning behind Verizon Wireless’ consolidation of 10 data centers to just three. The company’s CIO, John Hinshaw, has publicly claimed to have saved some $20 million in the process.
So for CIOs who want to “go green,” what are the best ways to get started? Here are some best practices, culled from experts in the field.
- Implement server virtualization to achieve greater efficiency in the data center. Virtualization software uses partitioning to create multiple virtual servers in a single hardware box, driving higher utilization rates and promoting server consolidation. For example, the utility Pacific Gas and Electric Co. — its territory includes California’s Silicon Valley—offers rebates of $150 to $300 for each server that companies eliminate using virtualization.
- Model energy consumption. Gartner’s Kumar recommends that CIOs invest both time and money to obtain what he calls “an accurate financial and technical handle” on their organizations’ data-center energy consumption. CIOs can run scenarios to assess the impact of sharp increases in energy costs, such as those seen in recent years. Kumar recommends a three-step process: consolidate, virtualize and then decommission equipment.
- Throttle back power consumption. IBM’s Anderson asks: “Why should servers be allowed to run at full power when they are idle?” His answer: Move the work to a second server—one with excess capacity—and throttle down or shut off the first server. IBM’s Active Energy Management technology, for example, enables processors to be shut down without disrupting the entire system, although the technology is not yet available on midrange or PC(including rack) servers. This capability is also used by organizations seeking ways to manage production spikes and disaster recovery, adds Anderson.
- Cool it, if you can. Some data-center managers hire design and assessment service providers to lighten their center’s power and energy consumption. Using a kind of Feng Shui for data centers, these consultants change the configuration of servers and equipment in the data center to ease air flow and allow refrigerated air to do its job more effectively.
- Get smart about storage. CIOs may want to check out MAID (massive arrays of idle disks), which is storage on cool — rather than hot — standby. MAID storage holds promise for next-generation mainframe storage expansion. Also, software can play a vital role in mainframe storage optimization, especially when used for tape management and optimization. When data is stored on fewer tapes, it can lead to higher storage utilization, meaning less handling of media or less hardware used; this, in turn, can reduce a company’s need for electricity.
Another good idea is a new twist on outsourcing. Lawrence of The 451Group suggests that some companies may start moving data centers to locations that are cooler—thereby requiring less air-conditioning— or that offer lower-cost electric power. “It’s an almost inevitable outcome that people will outsource new applications to alleviate this problem,” he says.
While that may be a little too much for some corporations to swallow, amore practical- sounding energy optimization plan —one that delivers immediate returns on investment—should not require too much energy for CIOs to sell.
Rusty Weston is a San Francisco-based journalist who blogs for FastCompany magazine and MyGlobalCareer.com.
|
Go Green for More Green “Green” strategies require careful technology planning. Done right, a green hardware strategy can improve efficiency, cost effectiveness and system reliability.
One surprising focus of the new, green agenda is the mainframe. A basic rule of process management is that any system works best with the smallest number of parts. The mainframe is the environment with the smallest number of (admittedly large) parts. This can lead to higher mean time between failures (MTBF), lower mean time to repair (MTTR) and higher overall reliability.
In fact, a small number of large boxes running at high utilization can be greener than a large number of smaller boxes running at variable levels of utilization. It also is much easier to govern, secure and manage a smaller number of assets.
Each part of an IT infrastructure has its own support infrastructure, including power supply, cooling system, physical casing, mounting, interconnections, console and buffers. These all consume both power and space. So by using the smallest number of individual subsystems, CIOs can reduce latent overhead to the absolute minimum.
Mainframes can also run reliably while maintaining high system utilization. This means that wasted processor cycles are reduced, in some cases to nearly zero. What’s more, with today’s performance management software, including solutions offered by CA, mainframes can be automatically tuned to offer even higher levels of performance.
A green IT strategy can also lead to management improvements. When CIOs look to minimize the environmental footprint of a system, they start with a clear process model that allows any redundant process to be eliminated. In this way, potential security holes can be caught and fixed. This approach also allows many of the weaker logical interfaces to be optimized, simplified or removed, leading to even higher levels of reliability. —R.W |