Manage what you measure with clarity
Article by Intel Data Center Management Solutions senior application engineer Rami Radi.
Once considered a nice-to-have, data center energy efficiency and sustainability have become top-of-mind if not mandatory considerations among data center managers over the past few years. While Big Tech firms such as Amazon, Facebook, Google and Microsoft run their data center operations very efficiently and have committed to transitioning to 100% renewable energy, globally, data center power consumption amounts to approximately 416 terawatts, or roughly 3% of all electricity generated on the planet.
Worldwide, Internet traffic tripled between 2015 and 2019 and is expected to double by 2022, according to Cisco. Meanwhile, the proliferation of artificial intelligence (AI) applications among enterprises and the anticipated surge in the number of sensors expected to be deployed with the mainstream adoption of the Internet-of-Things (IoT) raises the questions of just how much more energy usage in data centers will increase, and what this growth will mean for the release of carbon emissions globally.
Concerns about sustainability and the environment have led to renewed focus on varying metrics that assess the energy efficiency of data centers and the emergence of new standards. So, let’s take look at the most important and newest metrics to measure, determine and validate the efficiency of the data center.
Many metrics, but much progress needed in data center efficiency
Perhaps the most well-known metric used to gauge the energy efficiency of a data center is Power Usage Effectiveness (PUE). Introduced by the Green Grid in 2007, PUE is a ratio of the amount of power needed to drive and cool the data center versus the power draw from the IT equipment in the facility.
For the mathematicians in our audience, such an equation looks like this: PUE = Total Facility Energy ÷ IT Equipment Energy
Expressed as a ratio, the overall efficiency of a given data center improves as the quotient decreases towards 1. Put another way, a lower PUE is better because it indicates a smaller amount of energy is used for operations other than running the processors. To counter both rising electricity costs and concerns about carbon dioxide emissions, hyperscale cloud companies and larger colocation facilities have achieved annual or design PUE figures between 1.1 and 1.4.
The Uptime Institute has tracked industry average PUE numbers at various intervals for more than a decade, and its Global Data Center Survey 2019 report found that for the first time, there was no recorded improvement in energy efficiency. In fact, the efficiency ratings deteriorated slightly, from an average PUE of 1.58 in 2018 to 1.67 in 2019. Possible explanations range from extreme temperatures in regions of the world that necessitated increased use of cooling; to data centers with higher density racks that may have caused cooling systems to work harder; to cooling inefficiencies due to poor layout of servers and data centers operating below their optimal designs.
Data Center Infrastructure Efficiency (DCiE) is another Green Grid energy efficiency benchmark that compares a data center’s infrastructure to its existing IT load. Expressed as a percentage, DCiE is calculated by dividing IT equipment power by total facility power. For example, if a data center’s total IT load is 94kW and its total facility load is 200kW, that data center yields an efficiency score of 47%.
Developed by the U.S. Green Building Council (USGBC), Leadership in Energy and Environmental Design (LEED) is the most widely used green building certification system in the world. LEED uses a set of rating systems to provide independent verification of a building’s green features and sustainability initiatives, offering building owner-operators and designers the framework and metrics to ensure their properties are environmentally responsible and use resources efficiently. In addition to data center energy efficiency, LEED considers indoor environmental quality, location, materials and water usage. Interestingly enough, only a small percentage of data centers worldwide achieve LEED certification.
Founded in 1894, ASHRAE (American Society of Heating, Refrigerating and Air-Conditioning Engineers), is a global society advancing human well-being through sustainable technology for the built environment. For the data center sector, ASHRAE publishes specific guidelines and standards for temperature and humidity operating ranges of IT equipment.
Additionally, there is the PAR4 Energy Efficiency Rating System, which was developed by data center software provider, Power Assure, and is tested by the Underwriters Laboratories (UL). PAR4 is unique in that it measures server power consumption in several meaningful and more nuanced ways, including monitoring idle power, peak power, and total utilization power. This approach is not only logical but timely, since underutilized servers are a primary source of energy waste in data centers, especially during periods of low application demand.
What these and other standards and metrics make clear is that there are many opportunities to reduce energy consumption in data centers and lower costs while maintaining a safe environment for IT equipment, and in the process, curtail carbon emissions.
But because you can’t manage what you don’t measure, software solutions such as Intel Data Center Manager (Intel DCM) are needed to enable facilities and IT professionals to intelligently assess energy consumption. These tools provide real-time power and thermal consumption data, giving data center managers the clarity needed to lower power usage, safely increase rack density and prolong operation during outages.
Moreover, many data center administrators seeking to prevent infrastructure failures will overcool their systems and waste valuable energy because they lack visibility into actual server temperatures and power consumption. By aggregating server-inlet temperature data into thermal maps, data center management software solutions allow IT staff to monitor the effectiveness of their cooling solutions and airflow design. Data center management solutions also help identify underused servers, which not only saves power but helps to identify opportunities for consolidation and virtualization.
According to a new Uptime Institute report, efforts to improve the energy efficiency of the mechanical and electrical infrastructure of the data center are producing only marginal improvements. Now is the time for IT staff and facilities administrators to turn to data center management solutions to gain clarity into power consumption, significantly reduce energy use and carbon footprint, and slash mounting energy bills.