EPA Climate Leadership Award Winner Says Pursuing Environmental Sustainability Is Good for Data Centers
The recently released “United States Data Center Energy Usage Report” revealed that although power consumption continues to increase in U.S. data centers, the rate of increase has slowed substantially since 2005. The report was a combined effort by researchers from the U.S. Department of Energy (DoE), Carnegie Mellon University, Northwestern University, Stanford University and the Ernest Orlando Lawrence Berkeley National Laboratory. IBM has been at the forefront of delivering more efficient data center products, operations and services that have contributed to these reductions in the growth of data center energy consumption.
IBM’s Jay Dietrich and Alain Dubost have both been involved in identifying the data center energy efficiency improvements that contributed to IBM attaining the Environmental Protection Agency (EPA) 2016 Climate Leadership Award for goal setting. They agree that pursuing environmental sustainability in the data center has been good for business.
IBM earned this year’s award for its third-generation GHG emissions reduction goal. Established in February 2015, the goal commits the company to reduce carbon dioxide (CO2) emissions associated with IBM’s energy consumption 35 percent by year-end 2020, against a base year of 2005 and adjusted for acquisitions and divestitures.
“What we have consistently found is that [by] paying attention to environmental issues — specifically, energy consumption and greenhouse gas emissions — we make our operations more efficient,” said Dietrich, who is a Distinguished Engineer for energy and climate stewardship for IBM. “We use our materials and the energy we purchase in a more efficient manner, so we get more work out of them. And it reduces our costs and makes us more competitive while reducing our operational CO2 emissions, enabling IBM to set an aggressive CO2 reduction goal.”
The ‘How’ of Environmental Sustainability
According to the DoE report, power consumption in U.S. data centers grew explosively between 2000 and 2005 at a rate of 90 percent. Growth slowed to 24 percent between 2005 and 2010 and dropped to 4 percent between 2010 and 2014.
Although multiple factors contributed to slower growth in power consumption, the biggest was the reduced growth in the quantity of operating servers within U.S. data centers. Between 2000 and 2005, server shipments increased an average of 15 percent each year. Shipment growth fell to a 5 percent annual increase between 2005 and 2010 and slowed to 3 percent between 2010 and 2014. The report predicted that server shipments will stay at a 3-percent annual growth rate through 2020, with power consumption growth holding steady at a 4-percent annual increase.
According to IBM, increased virtualization has driven better use of equipment while reducing the data center’s server energy consumption and CO2 emissions footprint.
“If my equipment’s operating at 10 percent on average and I move that up to 40 percent, I can quadruple the workload but only increase energy consumption by 10 or 15 percent,” Dietrich explained.
Making the full use of the server capabilities enables IBM to deliver more work per unit of energy consumed and reduce the CO2 emissions associated with data center operations. The energy use reduction, in turn, is critical to IBM’s ability to set an aggressive CO2 reduction goal.
As an example, one IBM client consolidated the work of 300 x86 servers onto one IBM Flex System chassis, which consisted of six S822L servers and three IBM Storwize V3700 storage systems. The update cut annual data center energy consumption by 480 megawatt-hours and slashed CO2 emissions by 275 metric tons.
Improvements in Cooling
Dubost, who is the program manager for IBM’s Global Data Center, works to cut data center cooling costs by leveraging newer energy-efficient equipment and maximizing the use of free-cooling, using the low outside air temperature to cool data center equipment. It is particularly challenging to adapt IBM’s existing data centers to better use free-cooling and reduce cooling loads.
“[Globally], we also raised our data centers’ operating temperature to have those operate closer to the outside ambient air conditions, and reduce the amount of cooling equipment running,” explained Dubost. “By optimizing our cooling delivery and temperature, we were able to multiply the number of days these facilities operate on free cooling. Seventy percent of those key data centers are now providing at 23 degree Celsius air to the IT servers inlets.”
“Constantly, we’re working on upgrading those to the new technology and trying to reduce the level of cooling, the level of energy spend, in order to support our operations,” he added. “So we adapt all of the cooling equipment and the IT equipment to make those data centers always more efficient. In addition, we are using real-time monitoring and analytics systems to optimize cooling delivery and reduce cooling energy consumption.”
The data center, he says, is already starting to manage its own cooling system at all times according to the load. This includes the specificity of the layout to achieve the best level of balance. It is now able to quickly take the most convenient response in case an air flow situation occurs, by choosing to increase or reduce the ventilating equipment in order to always deliver the best and most efficient approach to cooling.
According to Dietrich, better cooling strategies can cut data center energy consumption by as much as 10 percent. However, better equipment utilization — including making use of the different opportunities in storage, server and network systems — can cut consumption by as much as 40 percent.
Automated Management Systems
One of the biggest challenges of operating an energy-efficient data center is ensuring instantaneous availability of resources while minimizing energy consumption.
“For some of our major clients, a 10-minute loss of availability is worth millions of dollars,” Dietrich said. “A few hundred thousand dollars in energy saved over a year or two doesn’t begin to compensate for that. The use of real-time automation and monitoring systems optimize IBM’s services while insuring a high level of reliability and resliency.”
Dietrich said automated management systems help data center administrators examine the performance of their servers, storage and network equipment, maintaining reliability while optimizing performance and power consumption. He added there are increasingly sophisticated tools to look at how a server is performing, how its different components — memory, processor and storage — are being utilized and how everything can best be optimized.
Creating a Culture for Climate Leadership
The innovations recognized in the Climate Leadership Award don’t just come from the top of any organization. Success requires a strong partnership between those who plan and those who implement energy-efficiency initiatives in the data center.
Although experts like Dietrich and Dubost may set goals related to energy efficiency, IBM’s real estate group has enterprise-wide responsibility for energy procurement and conservation, and the GTS operations team is responsible for implementing server and storage consolidation projects.
“There’s a good partnership between the parts of the business that develop solutions in this space and the groups that would implement those solutions,” Dietrich said. “These groups have to collaborate to deliver energy conservation projects and CO2 emissions reductions.”
The ultimate test of whether an energy-efficient solution passes muster is how well it delivers measurable business value. At the same time, Dietrich said that promoting sustainability is the right thing for businesses to do.
“Businesses are very large consumers of materials and energy in their operations and in the services and products they provide,” he concluded. “When we take action to make those reductions, they can have a very large impact.”