Problem solve Get help with specific problems with your technologies, process and projects.

Four ways to save data center power

With data center power usage skyrocketing, there are steps you can take to control costs. These four steps will lead the way to a more efficient data center.

At a recent virtualization event in Washington, D.C., one of the speakers from the Environmental Protection Agency (EPA) discussed a new report analyzing the growth of data center power consumption. As part of the movement to "go green," the U.S. government will be stepping up its efforts to control data center power use, with mandates for government data centers as well as recommendations for commercial data centers.

Though there is no indication today of actual mandates for the commercial market, there will be more tax incentives and utility rebates for companies that are proactive in cutting energy usage. As part of the report, the EPA released some sobering information: In 2006, U.S. data centers used 60 billion kWh, or 1.5 % of the country's total power consumption. More significant is that consumption has doubled in the past five years and is predicted to double again over the next five, with an estimated cost of $7.4 billion in 2011.

The data center power issue is not news to your customers. With the increase in the number and power consumption of servers and other equipment, and the rising price of energy, data center power costs have skyrocketed. Fortunately, there are steps your customers can take to help reduce these costs:

1. Consolidate/virtualize

Although servers running at low utilization levels (10% utilized) use less power than servers running at higher utilization, running lots of underutilized servers requires more power than running a few well-utilized servers. Reducing the total number of servers through either simple consolidation (combining like applications on fewer servers) and/or server virtualization (running multiple virtual servers on one physical server) can save your customers significant money on data center power. Consolidation ratios vary by workload, but many shops are able to run between 10 and 30 virtual servers (or even more) on one physical server.

2. Upgrade to more efficient hardware

There have been many recent improvements in energy efficiency -- in chips, power supplies, cooling technologies and management software. As part of the server consolidation process, talk with your customers about considering an upgrade to more efficient hardware, which can reduce data center power requirements while increasing processing power. In addition, if an organization is implementing virtualization, the newer chips (e.g., Intel-VT and AMD-V) include virtualization hardware assists that can make the virtualization software perform more efficiently.

Blade servers are also worth discussing with your customers, despite their negative reputation relative to power. Although blades use more power per square foot due to their higher density, they have better power efficiency, server for server, than their rack counterparts, in part due to sharing resources within the blade chassis, which reduces the total number of components.

Relative to all servers, both AMD and Intel have made strides in power efficiency with multi-core processors and greatly improved CPU-power-per-watt ratios. In addition, power supplies have gotten more efficient, delivering a higher percentage of the power from the wall to the server. Cooling systems have improved as well, with more efficient air flow, fans, blowers and so on.

3. Improve data center power management

Most servers today come with some form of dynamic power management, which can power down unused portions of the server, and then power them back up again when needed, thus reducing overall power consumption. This feature is typically disabled when shipped from the manufacturer. Enabling it is an easy way to reduce your customers' power usage, though it may cause a slight delay if there is an instant surge in demand that requires powered-off components to be powered on.

The increasing integration of virtualization and power management will bring us closer to intelligent policy-based management, where workloads can be moved across servers based on utilization, and management software will automatically power down servers completely when workloads are lower and bring them up again as needed.

4. Review the infrastructure and facilities

Because a great deal of data center power is used for cooling, improving air flow efficiency will cut power usage. Experts can conduct an energy audit of the entire environment and will often identify physical changes that could result in significant power reductions. Examples include employing a hot-aisle/cold-aisle approach for equipment (to make sure the hot air from one rack doesn't go into the input of another rack); installing blanking filler panels within racks to separate the air flow; clearing out the space under raised floor areas for better air flow; and ensuring proper placement of CRACs (computer room air conditioners) to maximize return air flow.

Perhaps more complex than these technical issues are the organizational issues -- it is important to help your customers build a bridge between the IT staff and the infrastructure folks so that there is a top-level view of the costs in the data center, including both the hardware/software and the infrastructure/facilities/power. Once an organization realizes how much is being spent on data center power, it may change the way it looks at upgrading the IT infrastructure.

About the author
Barb Goldworm is president and chief analyst of Focus Consulting, a research, analyst and consulting firm focused on systems, software and storage. Barb has spent 30 years in various technical, marketing, senior management and industry analyst positions with IBM, Novell, StorageTek, Enterprise Management Associates and multiple successful startups. A frequent speaker, columnist, and author of numerous white papers and research studies, she recently released a book titled Blade Servers and Virtualization: Transforming Enterprise Computing While Cutting Costs, published by Wiley. In 2007, she chaired the Server Blade Summit on Blades and Virtualization, and she has been the keynote speaker at numerous events on both virtualization and blades. She previously created and chaired the network storage track of Interop, and has been one of the top three ranked analyst/knowledge expert speakers at Storage Networking World. Barb can be reached at

Dig Deeper on Server management, sales and installation

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.