Cooling

Proper cooling in the data center is becoming more problematic as rack power density continues to increase.  Server consolidation, virtualization, and HPC all contribute to rack power density in the attempt to improve energy and floor-space efficiencies.  However, as power densities increase per rack, the wasted heat output also increases. Therefore, the traditional raised-floor cooling approach (designed for 3kW/rack) is no longer suitable or efficient for racks with power densities approaching 30kW. Furthermore, improper cooling of your facility and equipment can result in unpredictable software/data anomalies and increased hardware failures.

 

As critical as the cooling of a data center is, designing and optimizing a cooling system is a complicated undertaking and a much researched topic. Yet, the following strategies can help to ensure that proper cooling is achieved by adhering to a few simple principles.  Simple improvements to your cooling system can then be realized, without oversizing your current cooling system and thereby driving up your capital and operational costs with unnecessary expenditures.

 

 

Improve your data center's power efficiency.

Improving a data center's power efficiency will greatly reduce the necessary cost to cool the facility.  Utilizing higher voltages and the latest UPS/PDUs will improve overall energy efficiencies and thereby decrease wasted heat output.  Remember that 1 Watt of power consumed requires 1 Watt of cooling.  So, understanding your heat/wattage output will also help you to configure the optimal and most efficient cooling solution. See the section on Power for a more detailed explanation about the benefits of using higher voltages and latest generation power components.

 

 

Ensure that your current cooling architecture has been optimized and can handle an increased thermal load.

 

• Verify that your current maximum cooling capacity is not exceeded by the power requirements of the IT equipment.
• Measure and monitor the room and rack temperatures, making sure that all rack temperatures are within the equipment's specified operational ranges.
• Measure and monitor CRAC supply and return temperatures.
• Make sure all cooling components (pumps, condensers, cooling lines, etc.) are operating correctly.
• Keep flooring and sub flooring clean and unobstructed.
• Keep rack and sub floor cabling organized, in order to prevent airflow disruptions.
• Make sure that ventilation tiles are arranged properly and are distributing airflow evenly.
• Perform/schedule regular cooling system maintenance.
• Use blanking panels on racks with vacant spaces.
• Use cable cutout grommets to seal flooring.
• Replace tiles that have unused cable cutouts with full tiles.
• Separate high-density racks (not always feasible with HPC systems) to allow for the borrowing and averaging of the cooling supply.
• Use airflow assisting devices as required.

 

Modify your current rack arrangements.

 

• Arrange racks so that cold front-sides face each other, while hot rear-sides also face each other - creating cold and hot aisles.
• Manage the hot aisles by placing CRAC units along hot aisles.
• Use containment roofs and doors on hot aisles, forcing the hot air to be cooled more efficiently.

 

 

Research and employ the latest scalable and modular cooling enhancements.

 

• Use row-based and cold-aisle/hot-aisle architecture.
• Investigate supplemental cooling technologies that can directly target hot zones and provide in rack cooling.
• Investigate cooling options that can be hung overhead, saving valuable floor space.
• Investigate modular cooling designs that will allow you to utilize plug-and-play scalability.
• Use scalable cooling modules to place cooling as close to the heat source, and/or to target known hot-zones.
  
• Investigate thermally-neutral, fully contained cooling modules for your HPC deployment.

 

Additional Considerations:

 

• Consider the use of "air side economizing" by using external cold air for cooling during the winter months.
• Eliminate extraneous/superfluous cooling equipment (e.g. fans) that do little to cool, but increase energy and heat loads.
  

 

Remember that some of the solutions described are less costly than others. Some of the more cost-beneficial cooling solutions are to ensure that your current system is fully optimized and serviced, while properly sealing the floor and adopting row-based/hot-aisle arrangements.  Be sure to visit the supplemental Cooling site in addition to our Facilities Estimator for further cooling information and HPC configuration/cost analysis.

 

Whether you require modular systems for close-to-the-source hot-zone targeting, or a fully contained thermally-neutral system, Aspen Systems can help you decide on the appropriate cooling architecture for your HPC system.