Reducing data center power consumption

Of course, not every organization has Google’s resources or economy of scale. Nor does every data center use and conserve energy in the same way, because they have both different content and different processing requirements for that content. But there are many, many options for improving data center efficiency.

Best practices

Google’s best practices for data center usage efficiency include the following:

• Measure and track your power utilization efficiency (PUE).
• Manage airflow to minimize hot spots and hot/cold mixing.
• Avoid overcooling below equipment specifications.
• Lower cooling costs with water-based evaporative cooling systems and, when possible, add outside “air-economizers” to hotter re-circulated indoor air.
• Reduce overhead with high-efficiency power supply and UPS systems.
• Invest in energy-efficient servers and data storage.

Source:
google.com/corporate/green/datacenters/best-practices.html

Akamai runs 61,000 servers in 71 countries (often housed in third-party data centers), delivering as much as 15 percent to 20 percent of the World Wide Web’s daily traffic of content and transactions. Akamai squeezed 29 percent more productivity from its servers in 2009 by identifying and rewriting inefficient software to use less CPU and I/O capacity per server. According to the company, code optimization yielded a 29 percent improvement in server productivity in 2009. Combined with complementary improvements in hardware, Akamai saw a 32 percent reduction in energy use on network traffic and intends to cut an additional 35 percent in 2010. 

Keeping it real: IBM and HP

By Hugh McKellar, KMWorld editor in chief

IBM’s approach to data center design is to provide the most cost-effective solutions while offering flexibility to meet clients’ unpredictable changes in IT demand—and to do it in a manner that is faster to deploy. Along with increased changes in technology rates and unpredictability in most businesses, its strategy allows the data center to adapt to change.

In the past few years, IBM has designed more than 400 data centers, from Boston to Bangalore, with a modular plug-and-play design. According to recent industry surveys, more than 80 percent of clients plan to adopt that approach, a drastic increase from a reported 11 percent only two years ago.

The traditional design tends to include a large, monolithic data center build that is filled over time as the capacity needs of the business increase—effective when technology power and cooling requirements remained relatively stable and business growth was predictable. IBM emphasizes modular, plug-and-play design that is more cost-effective and adaptable by allowing clients to build in smaller increments, pay as they grow and add capacity when needed—thus deferring as much as 40 percent to 50 percent of the capital and operating costs.

As energy costs continue to increase by 10 to 25 percent per year, IBM’s designs have helped saved clients 20 to 40 percent per year in reduced energy costs. As power and cooling constraints in data centers remain the top concern with clients, each new generation of IBM servers is designed to provide better performance per watt—improving the performance for the business applications while keeping the power footprint flat.

HP’s application suite

HP reports that by automating data center management across life cycles, an IT organization can deliver cost-efficiency, quality and compliance in the most complex environments. Its Data Center Automation Center is a comprehensive suite of applications that focus on business requirements and produce business outcomes.

HP elaborates by saying its Data Center Automation Center can help organizations:

• reduce cost by automating day-to-day operations and shift funding to strategic initiatives,
• help ensure compliance by automating responses to Sarbanes-Oxley audits and similar regulatory requirements,
• standardize to optimize uptime and meet service levels using an up-to-date model of computing infrastructure with visibility down to individual software configurations, and
• increase availability by following best practices for change and configuration management and by standardizing builds and configurations at the time of deployment.

HP claims to have long recognized the importance of energy efficiency in its own operations and in product performance, and has now widened the focus to include suppliers by encouraging them to improve energy efficiency and increase use of renewable energy sources. The program is being implemented through supplier engagement, reporting greenhouse gas (GHG) emissions, establishing reduction targets and building suppliers’ capability in that area.

In fact, the company conducted a survey and received responses on energy use and GHG emissions in product manufacturing for 2008 from suppliers representing 86 percent of its material and manufacturing spend (up from 81 percent for 2007). Aggregate carbon dioxide equivalent (CO2e) emissions were 4.1 million metric tons, which is more than twice the emissions from its own operations. (For clarification, says HP, those numbers are not directly comparable to the 2007 figure of 3.5 million metric tons CO2e because it covers 5 percent more of its spending and because HP’s revenues have increased 13.5 percent during the period.)

HP adds its suppliers have become increasingly committed to reducing energy and emissions. The number of suppliers calculating and disclosing their emissions increased by a third, two-thirds of suppliers reporting to HP have established GHG reduction goals, and about 20 percent are estimating the emissions of their own suppliers.

As a caveat, many factories supplying HP also supply other electronics brands, so the company says it is impractical to measure the energy used to make HP products separately from its suppliers’ other business. Therefore, the share of their energy consumption is allocated in proportion to the value of HP’s business in its suppliers’ annual revenue. That method has its limitations, and HP says it will continue to work with suppliers to improve data quality.