Data Center Carbon Emission Effectiveness Oxycom Fresh Air BV November 20th, 2015 Abstract Carbon Emission Effectiveness (CEE) has been proposed as a new metric similar to the Power Usage Effectiveness (PUE), but focussed on carbon footprint rather than on energy consumption. It also properly incorporates the impact of on-site water consumption in evaporative cooling systems. A typical data center equipped with evaporative cooling technology instead of conventional air conditioning technology is expected to improve its CEE from 1.33 to 1.04, hereby reducing the CO 2 emission for the cooling equipment by 89% and the CO 2 emission for the IT equipment and cooling equipment combined by 22%. The on-site and off-site water consumption for the cooling equipment is reduced by 28%. For the cooling equipment and IT equipment combined, the water consumption reduction is 7%. 1 Introduction Data centers spend major part of their energy consumption on the IT equipment and on the cooling of the IT equipment. Over the last decade several metrics have been defined to represent the energy effectiveness of data centers. These numbers can be used as a reference in comparing and attempting to reduce their energy consumption and corresponding carbon footprint. However, the most commonly used metrics are primarily focussed on conventional air conditioning technology rather than on more energy-efficient evaporative cooling technology. In this article, a new metric will be proposed, compatible with both the old and the new cooling technology. Oxycom Fresh Air BV - P.O. Box 212, NL-8100 AE Raalte - Phone: +31(0)572 349 400 - E-mail: info@oxy-com.com 2 Data center cooling 2.1 Conventional air conditioning Conventional air conditioning technology is the most commonly used method to keep data centers cool, e.g. using chillers or direct expansion (DX) units. These systems are designed to recirculate the heated air returning from the server racks and cool it down to typically as low as 18-21 C [1], which is then supplied back to the server racks again. 2.2 Intel Already in 2008 Intel performed a comparative data center cooling test between conventional recirculating air conditioning and free cooling with outdoor air [2]. Whereas the conventional air conditioner cooled the returning air from the server racks down to a constant 20 C, the supply temperature with free cooling varied between about 18 C and 33 C with an absolute humidity equal to that of the outdoor air. The 10-month endurance test showed no significant difference in failure rate between identical data centers equipped with the different cooling methods. 2.3 Evaporative cooling State-of-the-art direct evaporative cooling technology is capable of achieving at least 90% saturation efficiency [3], hereby cooling air close to its wet bulb temperature while simultaneously increasing its moisture content. Indirect evaporative coolers may cool air even down to below its initial wet bulb temperature, without adding any moisture. Taking into account the results of the Intel data center endurance test and the fact that the worldwide wet bulb temperature seldom exceeds 30 C [4], evaporative cooling technology may be applied to data centers almost irrespective of the local climate type. Large-scale reverting to evaporative data center cooling has a massive 1
potential of reducing the ever-increasing cooling demand for data centers throughout the world. 3 Data center effectiveness 3.1 Introduction There are several quantities specially defined to describe to what extent energy in a data center is actually used for the IT equipment, such as servers, monitors and network equipment. 3.2 PUE The Power Usage Effectiveness (PUE) was introduced by The Green Grid in 2007 [5] and has become the most commonly used metric for expressing the energy efficiency of a data center. It is defined as: P UE = T otal F acility Energy [] (1) IT Equipment Energy includes all energy associated with the IT equipment. Total Facility Energy includes IT Equipment Energy as well as all other energy-consuming components, such as cooling equipment and lighting. In practice, PUE is always higher than 1; a lower value indicates a more energy-efficient data center. According to the Uptime Institute [6], its Network members showed a decrease in self-reported PUE for their largest site from 1.89 in 2011 to 1.7 in 2014, meaning that even in the best case still 41% of the energy was being used for other purposes than for the IT equipment. The PUE definition works well for data centers equipped with conventional air conditioning technology, since all cooling power originates from electricity-driven equipment. However, it is not applicable for data centers equipped with evaporative cooling technology. Although then the energy consumption for cooling is greatly reduced, hereby lowering the PUE value, the significant on-site water consumption is being ignored. Therefore, the PUE for data centers equipped with evaporative cooling systems is thought to give a falsely optimistic view. 3.3 CUE To focus more on data center sustainability, The Green Grid proposed the Carbon Usage Effectiveness (CUE) in 2010 [7]. It is defined as: CUE [ ] kg CO2 = T otal CO 2 Emission T otal Data Centre [kg CO 2] (2) CUE is related to the PUE metric through the Carbon Emission Factor (CEF): CUE [ ] kg CO2 = CEF [ ] kg CO2 P UE = (3) CO 2 emitted [kg CO 2] Unit Of Energy [] P UE However, it has some limitations: CUE is not a dimensionless quantity, which makes it difficult to compare it directly with the dimensionless PUE value. The numerator in this equation is directly derived from the numerator in the PUE definition. As such, on-site water consumption from evaporative cooling systems is still not included. 3.4 WUE In 2011 The Green Grid proposed the Water Usage Effectiveness (WUE) [8] in addition to the data center energy (PUE) and carbon (CUE) sustainability metrics. Furthermore, The Green Grid makes a distinction between direct on-site ("site-based") and indirect off-site ("source-based") WUE metrics: W UE [ ] L Annual Site W ater Usage [L] = W UE source [ L ] = Annual Source Energy W ater Usage [L] Annual Site W ater Usage [L] + (4) (5) Annual Site Water Usage refers to the on-site water use for the operation of the data center. Annual Source Energy Water Usage refers to the offsite water use for the generation of electricity. 2
The first term of the WUE source definition is related to the PUE metric through the Energy Water Intensity Factor (EWIF) that is based on the water used for the generation of electricity: [ W UE L ] [ source = EW IF L ] P UE + Annual Site W ater Usage [L] (6) Although the on-site water consumption of evaporative cooling systems is included in both the WUE and WUE source metrics, the fact that they are not dimensionless quantities implies the same limitations as the CUE metric. 3.5 WEE In 2015 Jack Pouchet from Emerson Network Power proposed the Water Equivalent Energy (WEE) [9] metric to be added to the numerator of the existing PUE definition in an attempt to incorporate on-site water consumption for data centers equipped with evaporative cooling systems. It is defined as: W EE [] = Actual W ater Consumed [L] W EEF [ ] (7) L The Water Equivalent Energy Factor (WEEF) represents the evaporative energy of water at 100 C (2256 kj/kg [10] or 0.627 kwh/l). The WEE, then, represents a fictional electrical equivalent of the cooling energy obtained from the evaporation of water. The flaw in this definition is that it falsely suggests that for example 1 kwh of electrical energy put into a conventional air conditioning system and the evaporation of 1.6 L of water would result in the same energy costs and environmental impact, whereas in this ratio water is both less expensive to purchase and more environmentally friendly to produce than electricity. 4 Carbon Emission Effectiveness 4.1 Introduction Despite several attempts to improve and expand the wellknown and widely used PUE definition, none seem to have succeeded in properly incorporating the impact of evaporative cooling. 4.2 Criteria A good data center efficiency definition should incorporate the following elements: It should be a dimensionless quantity, to allow a direct numerical comparison with the PUE metric. It should incorporate water consumption, both onsite (for the evaporation cooling) and off-site (for the generation of electricity). Conversion of the energy consumption and water consumption into their equivalent CO 2 emission rates allows for a direct comparison between the two quantities. Conversion of the energy consumption and water consumption into their equivalent CO 2 emission rates allows for an efficiency definition based on the environmental impact rather than on operating costs. Conversion of the energy consumption and water consumption into their equivalent CO 2 emission rates implies incorporating the specific fuel source used for the generation of electricity. 4.3 Definition Equivalent to the PUE definition, it makes sense to define a new dimensionless Carbon Emission Effectiveness (CEE) metric: CEE = T otal CO 2 Emission [kg CO 2 ] IT Equipment CO 2 Emission [kg CO 2 ] (8) IT Equipment CO 2 Emission includes all CO 2 emission associated with powering the IT equipment, such as servers, monitors and network equipment. Total CO 2 Emission includes all CO 2 emission associated with powering the IT Equipment and all other energy-consuming components, as well as onsite and off-site water consumption. 3
5 Calculation example 5.1 Introduction A calculation example of a typical 1 MW data center is shown below, including a comparison of the impact of conventional air conditioning versus evaporative cooling on the PUE and CEE values. 5.2 Assumptions Data center High Performance Computing (HPC) power: 1 MW (example value). COP of the conventional air conditioning system: 3.0 (typical value). All energy in the data center is used by either the IT equipment or the cooling system. Other sources are assumed to have a negligible impact. Average water consumption for the thermoelectric generation of electricity (coal, oil, natural gas, nuclear): about 4.16 m 3 /MWh [11]. Electricity consumption of evaporative cooling: 0.032 MW per MW HPC power [11]. Water consumption of evaporative cooling: about 7570 m 3 /year per MW HPC power [11]. Average CO 2 emission for the thermoelectric generation of electricity: about 0.75 kg/kwh [12]. Electricity consumption for the desalination of water: about 4 kwh/m 3 [13]. 5.3 Cooling analysis for a data center equipped with conventional air conditioning Air conditioning equipment Electricity consumption: 1 MW / 3.0 = 0.33 MW. electricity: 0.33 MW 4.16 m 3 /MWh 8760 h/year = 12 157 m 3 /year. CO 2 emission for the generation of electricity: 0.75 kg/kwh 0.33 MW 1000 kw/mw 8760 h/year / 1000 kg/tonne = 2190 tonne/year. IT equipment Electricity consumption: 1 MW. electricity: 1 MW 4.16 m 3 /MWh 8760 h/year = 36 472 m 3 /year. CO 2 emission for the generation of electricity: 0.75 kg/kwh 1 MW 1000 kw/mw 8760 h/year / 1000 kg/tonne = 6570 tonne/year. Total Electricity consumption: (0.33 + 1) MW = 1.33 MW. Off-site water consumption: (12 157 + 36 472) m 3 /year = 48 630 m 3 /year. CO 2 emission: (2190 + 6570) tonne/year = 8760 tonne/year. PUE: (1 + 0.33) MW / 1 MW = 1.33. CEE: (6570 + 2190) tonne/year / 6570 tonne/year = 1.33. 5.4 Cooling analysis for a data center equipped with evaporative cooling Evaporative cooling equipment On-site electricity consumption: 0.032 MW/MW 1 MW = 0.032 MW. electricity: 0.032 MW 4.16 m 3 /MWh 8760 h/year = 1160 m 3 /year. On-site water consumption for evaporative cooling: 7570 m 3 /(year MW) 1 MW = 7570 m 3 /year. Off-site electricity consumption for the desalination of water to evaporate on-site: 4 kwh/m 3 7570 m 3 /year / (1000 kw/mw 8760 h/year) = 0.003 MW. Combined (on-site and off-site) electricity consumption: (0.032 + 0.003) MW = 0.035 MW. Combined (on-site and off-site) water consumption: (1160 + 7570) m 3 /year = 8730 m 3 /year. Combined CO 2 emission: 0.75 kg/kwh 0.035 MW 1000 kw/mw 8760 h/year / 1000 kg/tonne = 232 tonne/year. 4
IT equipment Total CO2 emission reduction: 22% (from 8760 tonne/year to 6802 tonne/year). Electricity consumption: 1 MW. electricity: 1 MW 4.16 m3 /MWh 8760 h/year Figure 1 Power consumption = 36472 m3 /year. CO2 emission for the generation of electricity: 0.75 kg/kwh 1 MW 1000 kw/mw 8760 h/year / 1000 kg/tonne = 6570 tonne/year. Total Electricity consumption: (0.035 + 1) MW = 1.035 MW. On-site and off-site water consumption: (1160 + 7570 + 36472) m3 /year = 45 202 m3 /year. Figure 2 Water consumption CO2 emission: (232 + 6570) tonne/year = 6802 tonne/year. PUE: (1 + 0.032) MW / 1 MW = 1.03. CEE: (232 + 6570) tonne/year / 6570 tonne/year = 1.04. 5.5 Comparison when using evaporative cooling instead of conventional air conditioning (cooling equipment only) Total power consumption reduction: 89% (from 0.33 MW to 0.035 MW). Figure 3 CO2 emission Total (on-site and off-site) water consumption reduction: 28% (from 12157 m3 /year to 8730 m3 /year). Total CO2 emission reduction: 89% (from 2190 tonne/year to 232 tonne/year). 5.6 Comparison when using evaporative cooling instead of conventional air conditioning (cooling equipment and IT equipment) Total power consumption reduction: 22% (from 1.33 MW to 1.035 MW). Total (on-site and off-site) water consumption reduction: 7% (from 48 630 m3 /year to 45 202 m3 /year). 5
Figure 4 PUE and CEE values 6 Conclusion The example data center equipped with evaporative cooling technology shows very similar PUE and CEE values. This is caused by the fact that the CO 2 emission originating from the generation of electricity is greatly reduced, while the CO 2 emission originating from the desalination of water to evaporate on-site only has a modest impact. Although numerically the CEE metric is not expected to differ much from the PUE metric for data centers equipped with evaporative cooling technology, it will clearly show that switching from electricity to water as the power source to satisfy the cooling demand will greatly reduce the carbon footprint. A calculation example of a typical 1 MW data center has shown that the energy consumption and corresponding CO 2 emission rate for the cooling equipment only can be reduced by 89% when conventional air conditioning technology would be replaced by evaporative cooling technology. Including the IT equipment, the total energy consumption and corresponding CO 2 emission rate is reduced by 22%. Regarding the cooling equipment only, replacing conventional air conditioning technology with evaporative cooling technology actually reduces the water consumption by 28%, albeit that the majority will then be on-site rather than off-site. Including the IT equipment, the total water consumption is still reduced by 7%. Replacing conventional air conditioning technology with evaporative cooling technology improves the PUE from 1.33 to 1.03, meaning that only 3.1% of the total energy consumption is spend on non-it equipment instead of 25%. The corresponding CEE improves from 1.33 to 1.04, meaning that only 3.4% of the total CO 2 emission originates from non-it equipment instead of 25%. 6
References [1] Intel, The Efficient Datacenter Improving Datacenter Efficiency Through Intel Technologies and High Ambient Temperature Operation. [13] DESWARE, Encyclopedia of Desalination and Water Resources, Energy Requirements of Desalination Processes, http://www.desware.net/desa4.aspx. [2] Intel Information Technology, Reducing Data Center Cost with an Air Economizer (2008). [3] American Society of Heating, Refrigerating and Air-Conditioning Engineers, 2007 ASHRAE Handbook HVAC Applications (SI), Chapter 51 Evaporative Cooling (2007). [4] American Society of Heating, Refrigerating and Air-Conditioning Engineers, 2009 ASHRAE Handbook Fundamentals (SI), Chapter 14 Climatic Design Information (2009). [5] The Green Grid, PUE TM : A Comprehensive Examination Of The Metric, White Paper #49 (2012). [6] Uptime Institute, 2014 Data Center Industry Survey, https://journal.uptimeinstitute.com/2014- data-center-industry-survey/. [7] The Green Grid, Carbon Usage Effectiveness (CUE): A Green Grid Data Center Sustainability Metric, White Paper #32 (2010). [8] The Green Grid, Water Usage Effectiveness (WUE TM ): A Green Grid Data Center Sustainability Metric, White Paper #35 (2011). [9] Jack Pouchet, Power Usage Effectiveness Revised Part Two, Accounting for Water Use, http://blog.emersonnetworkpower.com/tag/waterequivalent-energy/. [10] American Society of Heating, Refrigerating and Air-Conditioning Engineers, 2009 ASHRAE Handbook Fundamentals (SI), Chapter 1 Psychrometrics (2009). [11] Otto Van Geet, National Renewable Energy Laboratory, Trends in Data Center Design ASHRAE Leads the Way to Large Energy Savings, NREL/PR-6A40-58902, http://www.nrel.gov/docs/fy13osti/58902.pdf. [12] U.S. Energy Information Administration, How much carbon dioxide is produced per kilowatthour when generating electricity with fossil fuels?, http://www.eia.gov/tools/faqs/faq.cfm?id=74&t=11. 7