Ashrae Journal - November 2008 - (Page 15)

cooling, lighting, and building switchgear, and account for 48% of consumption. The supply-side equipment is not an independent consumer of power; its power consumption depends on the power demand. Note that all data centers are different and savings potential will vary by facility. However, at minimum, this analysis provides an order-of-magnitude comparison for data center energy reduction strategies. The distinction between demand and supply power consumption is valuable because reductions in demand-side energy use cascade through the supply side. For example, in the 5,000 ft2 (465 m2) data center model, (referred to as “the data center model” in the rest of the article.) a 1 W reduction at the server component level (processor, memory, hard disk, etc.) results in an additional 1.84 W savings in the power supply, power distribution system, UPS system, cooling system and building entrance switchgear, and medium voltage transformer with no further action (Figure 1). Consequently, every watt of savings that can be achieved on the processor level creates a total of 2.84 watts of savings for the facility. 1. Processor power. In the absence of a true standard measure of processor efficiency comparable to the fuel efficiency standard for automobiles, thermal design power (TDP) serves as a proxy for server power consumption. The typical TDP of processors in use today is between 80 and 103 W (91 W average). For a price premium, processor manufacturers provide lower voltage versions of their processors that consume, on average, 30 W less than standard processors. Independent research studies show these lower power processors deliver the same compute performance as higher power models.4 In the data center model mentioned earlier, low power processors create a 10% reduction in overall data center power consumption. 2. Power supplies. As with processors, many of the server power supplies in use today are operating at efficiencies below what is currently available. The U.S. EPA estimated the average efficiency of installed server power supplies at 72% in 2005.1 In the data center model, it is assumed the unoptimized data center uses power supplies that average The Cascade Effect 79% across a mix of servers Recognition of the energy that range from four years old cascade concept leads to an to new. approach to reduce data center Best-in-class power supenergy consumption that starts plies are available today that at the core of the data center— deliver efficiency of 90%. the server—and follows the Figure 1:With the cascade ef a 1 W savings at the server component Use of these power supplies fect, flow of energy systematically level creates a reduction in facility energy consumption of 2.84 W. reduces power draw within to each subsequent level of the data center by 124 kW equipment. This enables compounding benefits throughout the or 11% of the 1,127 kW total. As with other data center data center. This Energy Logic approach also removes the three systems, server power supply efficiency varies depending most critical constraints faced by data center managers today: on load. Some power supplies perform better at partial loads than others and this is particularly important in dual-corded power, cooling, and space. Essentially, this sequential approach is the lever that enables devices where power supply utilization can average less than data center stakeholders to raise data centers to new levels of 30%. Figure 2 shows power supply efficiencies at different performance. It can deliver a 50% or greater reduction in data loads for two power supply models. At 20% load, Model A center energy consumption without compromising performance has an efficiency of approximately 88%, while Model B has or availability. When 10 top energy-saving strategies are ap- an efficiency closer to 82%. plied, in the order in which they have the greatest impact, to Another opportunity to reduce energy consumption is to the data center model, the strategies make available two-thirds size power supplies closer to actual server load. Server manuof floor space, one-third of UPS capacity, and 40% of precision facturers size the power supply for the maximum possible cooling capacity. server configuration, while the servers may be shipped out While the sequence is important, it is not necessary that with considerably less power requirement. Most servers do not each step can only be undertaken after the previous one is get upgraded to the maximum configuration; so these power complete. The energy-saving measures included in this se- supplies operate at small part loads, at low efficiency levels. quential approach to reducing energy consumption should End users should ask server manufacturers to size their power be considered a guide. Many organizations will already have supplies for the configuration they intend to use. undertaken some measures at the end of the sequence or will 3. Pow er manag ement softw are. Data centers are sized have to deploy some technologies out of sequence to remove for peak conditions that may rarely exist. In a typical business existing constraints to growth. data center, daily demand progressively increases from about November 2008 ASHRAE Journal 15 10 Energy-Saving Strategies

Table of Contents Feed for the Digital Edition of Ashrae Journal - November 2008

Ashrae Journal - November 2008 Contents Commentary Industry News Meetings and Shows Feature Articles Reducing Data Center Energy Consumption Building Sciences Feature Articles Sustainable Nature Center Emerging Technologies Special Products Washington Report Special Supplement: BACnet® Today BACnet® Goes Green Seeing the Light with BACnet® BACnet® on Campus BACnet® for Net Zero Analyzing BACnet® BACnet® at Xerox Campus BACnet® in Europe BACnet® for Green Library BACnet® at the University of Minnesota: Flexibility, Interoperability InfoCenter Product Showplace Classified Advertising Advertising Index

Ashrae Journal - November 2008

For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.