Data Center Journal

VOLUME 42 | FEBRUARY 2016

Issue link: http://cp.revolio.com/i/634770

Contents of this Issue

Navigation

Page 19 of 24

THE DATA CENTER JOURNAL | 17 www.datacenterjournal.com work harder (and therefore consume more energy) to move air through the system. Carlile added, "Reducing ΔP is the greatest way to reduce fan energy cost in a free- cooling data center. Since air filters can contribute significantly to the fan energy, reducing the ΔP of the filter will thus re- duce the overall fan energy cost." Because energy costs owing to pressure drop are so high, he recommends eliminating the typical prefilter when possible. counTing The cosT Clearly, free cooling offers potentially tremendous returns, particularly for data centers with high PUEs. e continued need for fans and filtration chip away at the benefits, but forgoing mechanical cooling (at least for the majority of the time) still yields tremendous operational savings. e question then comes down to the cost of implementation. Specifically, the two standard cases are designing a new facility with free cooling and retrofitting an older facility. Unfortunately for existing data centers, the cost and complexity of this project may make the rewards pale in comparison. "Retrofitting a data center to allow for free cooling is neither common nor easy nor inexpensive. In almost all cases, a data center is designed to allow for free cooling from the start," said Carlile. Of course, some organizations may wish to attempt such a challenging feat, but given already tight budgets at many companies, making the case for a difficult retrofit will be tough. at leaves new construction as the primary candidate for free cooling. One option, which is a possibility in most regions, is a design that implements free cooling exclusively. e other option provides some supplemental mechanical infrastructure to avoid temperature and humidity extremes or overly fast transi- tions (which can, for instance, lead to con- densation). Carlile notes that for the latter case, the two systems can be designed to avoid interfering with one another, par- ticularly with regard to filtration. "ese two technologies are designed to work in tandem at data centers that don't operate on 100% free cooling. In most cases, the equipment and/or building has louvers to control the percentage of outside air at any given time," he said. "In free-cooling mode, the building makeup air comes in via the outside-air louvers, and the building is vented either naturally or via pressure-relieving fans." When "full-blast" outside air is unsuitable, the system can automatically reduce the intake of outside air, turn down or deactivate pressure- relieving fans, and/or recirculate inside air. Naturally, the cost of implementing such a system varies by data center size and layout, equipment power density (in some cases, such as very high-density deployments, air is simply insufficient to provide the required cooling), and so on. When it comes to filtration, Carlile said that 24" by 24" (nominal) is the most com- mon filter type, allowing 492–500 linear feet per minute (FPM) of airflow and handling up to about 2,000 cubic feet per minute (CFM). He suggests a budget of about $50–$150 annually for each of these filter slots. According to 3M data, energy cost per filter rises approximately linearly with pressure drop. At $0.10 per kilowatt-hour, a ∆P of 0.25" (about 60 pascal) costs about $86 annually. At a ∆P of 2" (an 8x increase, or about 500 pascal), the energy cost per filter spikes to about $685 per year, illustrating the benefit of filters that do their job while minimally obstructing airflow. Since pressure is force per unit area, o ne option for reducing ∆P is to increase the area; Carlile also notes that data center mechanical engineers are "oversizing" air-filter racks/walls. e equivalent in our analogy might be a larger straw to provide more surface area for air to pass. e result would be less effort in moving air, meaning lower energy consumption. Filter costs may change somewhat given the size difference, but the change would likely be negligible. e critical takeaway from the effect of ΔP is that the use of energy-saving fil- ters is a simple means of reducing cooling costs without necessarily making changes to the data center infrastructure (although, as mentioned above, some relatively minor changes can enhance the benefit). Given a large number of filter slots in the data center, the energy savings from such filters can be quite large, potentially far exceed- ing the initial cost difference compared with higher-ΔP alternatives. Furthermore, if the filter replacements enable operation without prefiltration, the company can cut out an extraneous stage (and a source of ∆P) to gain additional savings. concLusions ASHRAE's latest guidelines make free cooling an option for almost any data center. One of the critical tasks for a successful implementation is filtering the outside air to prevent contaminants from wreaking havoc on electronic equipment. Unfortunately, however, filters make the air-distribution system work harder, in- creasing energy consumption. Two simple yet effective means of improving this situation are elimination of the prefiltering stage and use of appropriate low-pressure- drop filters. Of course, better filters cost more, so the key for the data center manager in making the case to the C-suite for the more-expensive options is to show how they will reduce the facility's total cost of ownership. is calculation, beyond just the initial cost of the filters themselves, includes the energy savings as well as the labor and disposal savings—particularly if the strategy includes elimination of prefilters. Free cooling enables tremendous energy savings, but the opportunity goes beyond just switching from mechanical air conditioning to outside air. A proper filtration strategy using low-∆P filters and eliminating unnecessary stages helps make free cooling even more beneficial to the data center. n

Articles in this issue

Links on this page

Archives of this issue

view archives of Data Center Journal - VOLUME 42 | FEBRUARY 2016