Document carbon footprint evaluation

Jürgen Knödlseder
1 parent ab3392f4
Showing 2 changed files with 64 additions and 0 deletions
doc/source/users/environment.rst
doc/source/users/index.rst
+.. _sec_environment:
+
+Environmental impact
+====================
+
+Like any human activity, using ctools will have an environmental impact. This
+includes for example the greenhouse gas emissions related to electricity
+production that is needed to run your computer or data centre. It includes as
+well the environmental footprint of your computing hardware, the power and
+material behind your internet connection, your storage array, and even the
+footprint of the personnel that eventually runs your data centre.
+
+As ctools user you should be aware of this footprint, and you should aim
+to reduce it as much as possible.
+
+To guide you in this endeavor, all ctools and cscripts inform you at the end
+of the log file about the estimated carbon footprint of your computations:
+
+.. code-block:: bash
+
+   2022-02-18T22:33:26: Application "ctlike" terminated after 3910.0 wall clock seconds, consuming 3881.18 seconds of CPU time and generating a carbon footprint of 3.81083 g eCO2.
+
+This estimate is based on the study of
+`Berthoud et al. (2020) <https://hal.archives-ouvertes.fr/hal-02549565v4/document>`_
+who assessed the carbon footprint of the DAHU cluster of the UMS GRICAD (Grenoble,
+France) in 2019. Their study included
+
+* server fabrication
+* server environment
+* server usage (electricity)
+* travelling of personnel in the context of their work
+* travelling of personnel from home to office
+* personnel equipment
+* personnel energy
+
+`Berthoud et al. (2020) <https://hal.archives-ouvertes.fr/hal-02549565v4/document>`_
+found that, per CPU hour, the footprint of the GRICAD data centre was 2.43 g CO2e
+(equivalent CO2 emissions) due to electricity use and 2.25 g CO2e due to the data
+centre environment, hardware fabrication and the personnel. For their study a
+carbon intensity of 108 g eCO2 / kWh for greenhouse gas emission due to
+electricity generation was assumed.
+
+The latter varies substantially between countries. While the current electricity mix
+in France has a (rounded) carbon intensity of 60 g eCO2 / kWh, values go through
+240 g eCO2 / kWh (Spain), 410 g eCO2 / kWh (Italy), 460 g eCO2 / kWh (Germany),
+520 g eCO2 / kWh (USA), up to 840 g eCO2 / kWh (Australia). So eventually, you should
+think about doing your computations in a country with a low carbon intensity, or make
+sure that your data centre is running on a low carbon energy source.
+
+ctools will determine in which country your code is executed and fetch the appropriate
+carbon intensity for the general energy mix in this country to estimate the carbon
+footprint of your computations. The estimate is done using
+
+.. math::
+   {\rm Carbon\,\,footprint} = \left( 2.25 + 2.43 \frac{\rm CI}{108} \right) \times {\rm CPU\,\,hours}
+
+where :math:`{\rm CI}` is the carbon intensity of the electricity mix in your country
+in units of g eCO2 / kWh, :math:`{\rm CPU\,\,hours}` is the number of CPU hours your
+computation took, and :math:`{\rm Carbon\,\,footprint}` is the carbon footprint in g eCO2.
+
+The carbon footprint of all ctools and cscript runs are stored in an ASCII file
+that you will find in a folder under your home directory, named ``.gamma/statistics.csv``.
+
@@ -14,4 +14,5 @@ This documentation will guide you in the process of using ctools.
    doxygen
    tips/index
    user_contributed/index
+   environment
    glossary