There have been some debate about the source of electricity a data center is using and the CO2 emissions it is causing.
Recently some interesting figures came available by the International Energy Agency. These are the CO2 emissions per kWh electricity generation. Published in the 2013 edition of “CO2 emissions from fuel combustion – Highlights”.
It isn’t easy to find consistent and complete time series. A lot of the data that can be found is using different definitions and/or different time periods what makes it difficult to aggregate these figures. IEA has published time series for the period 1990 – 2011.
To make some comparisons a selection from different parts of the world is showed in table 1. Remarkable differences in CO2 emissions can be found. Some countries show a huge decrease of CO2/kwH emission during the period 1990 – 2011 whereas others show an increase. Also within a region the differences are considerable. Zooming in on the E.U. countries with Tier 1 data center markets; United Kingdom, France, Germany and The Netherlands, (with the DC hubs London, Paris, Frankfurt and Amsterdam) we see a CO2/kwH reduction of 34.4%, 41.9%, 21.4% and 33.4%. Differences in emissions and emissions trends are caused by different energy policies and different compositions of the power plant fleet.
Table 1. CO2 emission per kWh from electricity generation, source IEA.
2011kg CO2 /kwH | Difference 1990 -2011
% |
|
E.U. | 0.352 | -21.4 |
United Kingdom | 0.441 | -34.4 |
France | 0.061 | -41.9 |
Germany | 0.477 | -21.4 |
The Netherlands | 0.404 | -33.4 |
Russian Federation | 0.437 | 7.6 |
U.S.A. | 0.503 | -13.6 |
Canada | 0.167 | -14.8 |
Australia | 0.823 | 0.7 |
Singapore | 0.500 | -44.9 |
Japan | 0.497 | 14.3 |
Korea | 0.545 | 4.8 |
India | 0.856 | 5.4 |
China | 0.764 | -14.5 |
The figures that are showed are averages. The CO2 emission of a data center depends on the power plants that are really used to deliver electricity to the data center. Depending on the electricity demand the power supplier will assign different power plants. The assignment of power plants is according to their production efficiencies (short-run marginal costs of production) and capacity and this production mix will influence the CO2 emission per kwH.
CO2 emission per server
To get an impression of the CO2 emission per server in different parts of the world we making use of the report ”Estimating total power consumption by servers in the U.S. and the world” of J.G. Koomey of Stanford University, the power usage of low, mid and high range server are estimated on 180, 420, and 4800 Watt. This will lead to the figures in table 2 based on a 24 hours x 365 days usage.
Table 2. Yearly CO2 emission per server.
Kg CO2/year | Low range server | Medium range server | High range server |
E.U. | 555 | 1295 | 14801 |
United Kingdom | 695 | 1623 | 18543 |
France | 96 | 224 | 2565 |
Germany | 752 | 1755 | 20057 |
The Netherlands | 637 | 1486 | 16987 |
Russian Federation | 689 | 1608 | 18375 |
U.S.A. | 793 | 1851 | 21150 |
Canada | 263 | 614 | 7022 |
Australia | 1298 | 3028 | 34606 |
Singapore | 788 | 1840 | 21024 |
Japan | 784 | 1829 | 20898 |
Korea | 859 | 2005 | 22916 |
India | 1350 | 3149 | 35993 |
China | 1205 | 2811 | 32125 |
Data center use case
What do all these figures mean for a data center? Lets take for example a data center of 1000 servers with a PUE of 1.8. In this case we use a server mix of 95% low range, 4% mid range and 1% high range servers. Besides servers the data center will also use storage and network components. The ratio of the energy use of servers versus the energy use of storage and network components is set to 75:15:10.
We can define a worst-case scenario when electricity is created with conventional coal combustion; in that case 1kW of electricity is equivalent to 1 kg CO2 emission. For the data center in this use case, that would be an upper limit of 4957 ton CO2 per year. In reality power suppliers are using a mix of different energy sources. As we can see in table 3, the lowest emission is 302 ton and the highest emission is 4244 ton. A difference with a factor 14!
Table 3. CO2 emission of a data center.
Metric ton CO2/year | Servers | Storage | Network | Data center |
E.U. | 727 | 145 | 97 | 1745 |
United Kingdom | 911 | 182 | 121 | 2186 |
France | 126 | 25 | 17 | 302 |
Germany | 985 | 197 | 131 | 2365 |
The Netherlands | 835 | 167 | 111 | 2003 |
Russian Federation | 903 | 181 | 120 | 2166 |
U.S.A. | 1039 | 208 | 139 | 2494 |
Canada | 345 | 69 | 46 | 828 |
Australia | 1700 | 340 | 227 | 4080 |
Singapore | 1033 | 207 | 138 | 2479 |
Japan | 1027 | 205 | 137 | 2464 |
Korea | 1126 | 225 | 150 | 2702 |
India | 1768 | 354 | 236 | 4244 |
China | 1578 | 316 | 210 | 3787 |
Zero emission
There is of course the alternative case of zero CO2 emissions if the electricity supply is completely based on nuclear, hydro or renewable energy. Some countries like Iceland, Norway, Sweden and Switzerland have extreme low CO2/kwH emission (1, 13, 17 and 30 gram).