Cutting Costly Data Center Energy Consumption
How real-time wireless monitoring pinpoints opportunities for efficiency gains
The GSA, which leases approximately 1,400 data centers located in large federal facilities to many government agencies, is under mandate to cut energy consumption in those facilities by 30 percent. In October 2011, the GSA’s STAR database for Automated Data Processing (ADP) space found that nearly half of data center energy is used for non-IT loads, such as cooling and power conditioning. The U.S. Environmental Protection Agency (EPA), in a 2007 Energy Star program report, predicted that GSA can expect data center energy use to grow at an annual rate of 15 percent, which represents a doubling of energy consumption every five years.
The U.S. Department of Energy (DOE) and the EPA recently sponsored studies that concluded that energy use can be reduced by 25 percent through the implementation of best practices and commercially available technologies. Clearly, data centers are a prime target for such energy-saving measures. Unfortunately, these centers are typically ‘buried’ in the large facilities where they’re housed, providing GSA with limited visibility into specifics around energy consumption. Access to such in-depth data would be invaluable in driving data center optimization. And by achieving greater energy efficiency, the GSA study, “Wireless Sensor Network for Improving the Energy Efficiency of Data Centers,” conducted by the Lawrence Berkeley National Laboratory in March 2012, concluded that the GSA stands to gain critical flexibility in future data center expansion planning through the reduction or elimination of additional power and cooling demands.
WIRELESS SENSOR TECHNOLOGY
Before GSA could proceed toward meeting its mandate to lower its energy and the associated greenhouse gas emissions, an energy consumption baseline had to be established. The GSA study’s authors, Rod Mahdavi and William Tschudi of Lawrence Berkeley National Laboratory (LBNL), reported that wireless sensor networking technology was chosen over wired sensors because of the high cost and complexity associated with hardwiring sensors in existing facilities. Plus, wireless systems are easily expandable and can be redeployed as needed for IT equipment refreshes, along with allowing them to meet future data center growth requirements.
The GSA, DOE, LBNL, and technology provider, SynapSense collaborated to evaluate the full potential that wireless sensor networking technology offers data center operators. To this end, GSA turned to its Green Proving Ground program (GPG) to conduct the study. Much as its name suggests, the GPG tests the efficacy of groundbreaking sustainable building technologies. The GPG program was created to draw upon emerging technologies in the private sector and drive improved environmental performance in federal buildings, using the GSA’s extensive real estate portfolio as a proving ground to help promote the implementation of exciting new breakthroughs.
The GPG study was aimed at validating how well the SynapSense Data Center Infrastructure Management Platform would provide data center operators with detailed, real-time measurements of environmental factors and power consumption. In this way, a performance baseline could be established, allowing operators to uncover areas of less-than-optimal performance, which in turn would present opportunities to reduce overall energy consumption in a way that is both cost-effective and facility-friendly.
The U.S. Department of Agriculture’s (USDA) National Information Technology Center (NITC) Data Center in St. Louis, was selected as a demonstration facility for the SynapSense wireless monitoring solution because of NITC’s determination to gain greater visibility into the operational efficiencies of its data center. The NITC team was intent on making improvements in both the energy efficiency and the overall resiliency of their facility. The team had learned about the SynapSense Data Center Optimization Platform during a presentation by Dale Sartor of Lawrence Berkeley National Labs at a 7x24 Exchange conference. This prompted the subsequent study, which focused on identifying ways to improve the efficiency of data center cooling.
Once installed, the network of wireless sensors measured floor-to-ceiling conditions, including humidity and temperature conditions at multiple elevations, the raised floor differential pressure, as well as UPS and CRAC cooling system power, which were used to calculate energy usage. Eighty humidity sensors measured the top of the cold inlet airside of each rack. Sixteen pressure sensors monitored the sub-floor to room differential pressure in the cold aisles. Sensors were also installed on the computer room air conditioner (CRAC) units to monitor supply and return air temperatures, along with relative humidity.
SynapSense’s comprehensive, integrated data management software analyzed the collected data, allowing data center operators to easily gauge facility performance in real-time. The data were then fed into the DOE’s assessment tool to calculate power usage effectiveness (PUE) with accurate power measurements
According to Mahdavi and Tschudi’s report, it was now possible to analyze the effectiveness of recirculation and bypass air mixing, under floor air pressure, as well as cooling system efficiency. Operators had the data they needed to determine if these critical systems properly adhered to thermal-operational ranges published by the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) for IT equipment. It was also possible to set up alerts in the event ranges were exceeded.
The data collected from the wireless sensors made it possible to analyze temperature, relative humidity, and dew point levels down to the individual racks. This analysis enabled operators to identify overcooling, overheating, and air mixing conditions in specific racks or aisles over time. Aggregation of real-time data from hundreds of sensors was then used to produce thermal imaging maps, which provided a visual snapshot of the environmental conditions in the data center. This made it far easier to rapidly pinpoint areas of concern.
The comprehensive data and resulting analysis made possible by the 600 SynapSense wireless sensors enabled the evaluation team to understand operating conditions and identify problem areas. Armed with this vital information, they could now explore ways to achieve desired efficiencies and energy savings.
After gaining authorization from USDA management, the team began implementing the data center changes called for by the in-depth analysis. One of the key findings was overcooling of the data center. Further investigation revealed that a high volume of air and low supply air temperatures the CRAC systems provided the racks was at the heart of the overcooling problem. The wireless system revealed that the vast majority of the racks were being operated below the ASHRAE-recommended temperature range.
Mahdavi and Tschudi’s report notes that maintaining temperature within the industry standard range is vital in order to minimize the possibility of thermal shock damage to equipment should a cooling outage occur that results in a >9.0°F elevation in temperature per hour. Running systems above or below the ASHRAE range for extended periods also increases the danger of equipment malfunction or failure. Beyond the risk to equipment, this practice also represents an inefficient use of cooling energy.
In dealing with the overcooling problem, the evaluation team tested several approaches and settled on a solution that shut down three CRAC units, then sealed them off so cold air couldn’t seep back into the data center through the pressurized, raised floor. Return air temperature set points were elevated and return air relative humidity set points lowered to the recommended settings. Once these steps were complete, the team was able to shut down the dehumidification and reheat modes on the CRACs, while demonstrating with real-time metrics that the facility met ASHRAE-recommended dew point range.
While overcooling was the predominant problem, the wireless sensor data also revealed some hot spots. Hot discharge air from the IT equipment was re-circulating back through the racks. This problem was traced back to a lack of blanking panels, which was easy to remedy.
Data gathered from wireless sensors that measure inlet air temperature were used to rebalance the cooling system. This was accomplished by examining the number and location of perforated floor tiles in the cold aisles and rebalancing the quantity and location of perforated and solid tiles where needed to match the IT cooling requirements of each rack. Solid tiles replaced 45 perforated floor tiles, which helped solve this problem.
Once these procedures were completed, the evaluation team spent the next two hours capturing and analyzing a new set of data to determine the impact of the changes made. Additional adjustments resulted in further fine-tuning of the systems. The entire implementation phase of this project was completed in less than a day. (See LiveImaging thermal, pressure, and humidity maps that detail the results of the optimization.)
COOLING SYSTEMS OPTIMIZED
The first set of results assessing the use of wireless sensor technology in data centers showed significant energy savings. Following the installation of SynapSense wireless sensor networking technology, subsequent data analysis and the resulting cooling system modifications, the NITC data center’s facilities staff was able to reduce cooling loads and achieve significant energy savings.
Mahdavi and Tschudi’s report concludes that the NITC data center was able to reduce cooling loads by 48 percent, thereby decreasing overall data center energy use by 17 percent. This represented an annual savings of 657 megawatt-hours (MWh). The report also notes a corresponding reduction in PUE, from 1.83 to 1.51.
The GPG study has confirmed the cost effectiveness of this innovative wireless technology. Despite the fact that this NITC facility already had some of the lowest energy costs of all of GSA’s facilities, calculations show that it will take less than three and a half years to offset costs of vendor provided hardware, software, and labor. The GSA report further concludes that applying this technology across the agency’s portfolio of tenant data centers would yield a potential savings of $61 million annually.
The substantial reduction in cooling load and CO2 emissions achieved as a result of GSA’s NITC data center study clearly demonstrates the efficacy of wireless sensor networking technology.
"By most standards, this data center is an efficient facility. The fact that a wireless sensor network helped it significantly reduce its energy profile speaks volumes for the technology," said Ron Jones, facility manager, Office of the Chief Information Officer, USDA.
The GSA study validates that wireless sensor networking technology can be instrumental in helping to achieve tremendous cost and energy savings, even at data center facilities which are well-designed and well-managed. The findings reported by Mahdavi and Tschudi suggest that wide spread adoption of this technology offers significant energy saving and environmental benefits that the broader data center industry simply can’t ignore.