How to cut data center energy use
Connecting state and local government leaders
Increasing density of infrastructure and system components will help agencies offset growth and reduce total energy usage.
In January, the House of Representatives unanimously passed for the Energy Efficient Government Technology Act (HR 306), which mandates that federal agencies develop plans for energy-efficient technology in their data centers. The Center for Climate and Energy Solutions found that widespread adoption of energy efficient information technologies could save the federal government more than $5 billion in energy costs through 2020.
Many government IT managers, however, may wonder how they can cut energy use when demand for data center services is growing.
One way agencies can meet the programs goals is to move activities to the public cloud, which would reduce direct energy usage by agencies. Public cloud providers employ the most advanced technologies in their high-density data centers to deliver the lowest operating cost and power usage per square foot. However, moving to a public cloud is not an option for all federal IT use cases.
Agencies that continue to run their own data centers must focus on reducing power and energy usage per square foot. To increase data and computing requirements without additional infrastructure space, agencies must increase density to offset growth and reduce total energy usage.
To do this, agencies should identify and implement best practices in the following areas:
Physical data center. Study the hyperscale data center firms that have pioneered best practices, including Facebook, Google and many others. The Open Compute Project provides a good blueprint for efficient, flexible and scalable hardware.
Infrastructure components. At the component level, density per square foot is the best metric to use in evaluating any infrastructure components. Currently available tools can help increase density and reduce power.
- Technology that actively measures energy usage by components in the rack enables data centers to limit their investment in power until needed. It also allows data centers to be more flexible in design and buildout, a good way to delay excess power usage.
- GPUs improve density per compute cycle dramatically and can be applied to great success for many high-performance computing problems.
- Switching infrastructure can be flattened by moving away from hierarchical leaf and spine switches and moving to optical switches that can reduce the overhead of ports that are used just for interconnecting switches.
- Helium storage drives reduce power consumption by up to 30 percent. In addition, there are new disk technologies, such as SMR drives, that can dramatically increase density for many use cases.
System-level infrastructure. Energy use can also be controlled with system-level components.
- Intelligently tier data onto the lowest cost, most dense storage infrastructure possible given data access requirements.
- Deploy scale-out storage and computing architectures that maximize utilization of infrastructure automatically. Today’s object storage solutions, for example, keep the overhead of data copies very low while allowing the efficient storage of massive-scale data.
- Use software intelligence to optimize all infrastructure to get the best results with the least infrastructure and the least power.
These best practices can help federal agencies meet the energy efficiency goals in the new Energy Efficient Government Technology Act. In fact, this bill can help federal agencies to adopt IT best practices at an accelerated rate, putting them on par with private companies.