cloudbasedneuroimaging / cloudbasedneuroimaging Goto Github PK

I noticed that the word cloud is "Cloud" in some places and "cloud" in others.

It would be interesting to think about administrative issues that can help/hinder the use of cloud computing. One example: the UW has a policy where F&A is waived for cloud computing. See: https://itconnect.uw.edu/research/waiver/

Privacy Shield used to be of interest for European researchers but unfortunately it has been declared invalid.

It would be good if information regarding the environmental cost of cloud computing were added to the section, Costs, that currently only discusses the financial burden. These environmental costs merit equal consideration given the seriousness of the ongoing climate emergency. Discussing the financial and environmental costs within the same section would be coherent as there is overlap in the steps to be taken to reduce each.

The operation of data centres contributes to an estimated 1% or more of global electricity consumption [1]. Much of that energy is produced from polluting sources, such as coal or gas. As the amount of computation being done increases, so too does the amount of energy required and thus the amount of greenhouse gases emitted (unless the energy is obtained from renewable sources).

Cloud computing providers can reduce their environmental impact by increasing the efficiency of their operations and sourcing more of their energy from renewable sources. Different providers do this to different degrees, with some committing to carbon neutral operations but others being less diligent. Amongst the large North American companies, Google and Apple (for example) perform well in this regard whilst Amazon performs relatively badly. Large East Asian providers, such as Alibaba and Baidu, lag even further behind [2].

The first way in which researchers can reduce their environmental costs is to compare different providers and select, wherever possible, the ones that have the least polluting operations.

More computation time means more energy and thus more pollution. This overlaps with the issue of financial cost, where more computation time means more money. The solution is the same for both costs - to reduce the amount of computation time.

The second way that researchers can reduce their environmental costs is to optimise their code and data so that it runs in the most efficient manner possible. Connected to this is thinking seriously about whether running a particular set of computations is strictly necessary and if the potential outcome outweighs the environmental (and financial) cost.

Adding this information would help make researchers more aware that their work necessarily has an environmental cost and would give them simple ways to help reduce that cost.

[1] https://www.nature.com/articles/d41586-018-06610-y
[2] http://www.clickclean.org/downloads/ClickClean2016%20HiRes.pdf