Adaptive Cooling in Data Centers

Chris Page, Yahoo!
Ray Pfeifer, Silicon Valley Leadership Group

This presentation will focus on a demonstration project for adaptive cooling of a production data center, as part of a research project sponsored by Silicon Valley Power's Energy Innovator Grant program. The field evaluation included specific characteristics of the existing data center air cooling systems, characteristics of server loads, and measurements for use of evaluating their operation performance under various operating conditions before an after adaptive cooling. The final case study will include the results for the proof of concept including:

• Technical objectives
• Technical information on the data center cooling systems and servers
• Measured parameters
• Test procedures and sequence
• Data acquisition and compiling process
• Performance metrics for adaptive cooling

To the extent possible, the case study will include monitoring results from the facility supporting the data center that housed the studies. In combination with the evaluation of energy performance of the test environment versus that of the balance of the facility, estimates of potential energy savings for expanding the adaptive cooling solution to the balance of the data center will be discussed and may be drawn from the data analyses in the report.

The project evaluated the SynapSense mesh wireless sensor network centralized adaptive cooling solution for a portion of the Yahoo! data center located in Santa Clara, California. The primary objective of the test was to demonstrate the effectiveness of the wireless sensor network and adaptive cooling alternatives for data centers, through performing on-site measurements and evaluation of the system performance. The technical objectives of this demonstration project included:

• Improve air flow management through installation of cold aisle containment.
• Demonstrate the ability to reduce cooling energy use by 25 percent (~1 megawatt (MW) for entire data center) or greater through the use of SynapSense control solution.
• Demonstrate the ability to power manage servers to reduce server power energy use and associated cooling energy by 40 percent (~2MW for entire data center).
• Quantify the potential carbon footprint reduction by reduced power generation emissions.

The following parameters were monitored or measured during the evaluation:

• Power demand of servers and computer room air handling (CRAH) cooling modules.
• Actual power demand for servers used in this study.
• Actual power demand for the cooling modules.
• Cold inlet air temperature to the server racks.
• Hot outlet air temperature from the server racks.
• CRAH cooling module inlet and outlet air temperatures.
• CRAH cooling module outlet relative humidity.
• Sub-floor/data center differential pressure.
• Total cooling tonnage provided by chiller plant.
• Power loss of UPS & PDUs serving servers.
• Outside air temperature and relative humidity.
• CRAH cooing efficiency (tons and kilowatts).
• CRAH air loss efficiency (percentage of wasted cooling cubic feet per minute).

The following parameters were monitored or measured during the evaluation to quantify the magnitude of power demand in the data center and chilled water plant:

• Total power demand going into all information techonology equipment.
• Total power demand to chiller plant.

Back to Agenda