Measure and manage the carbon footprint of cloud infrastructures and applications

First, a reminder: the cloud and its different service levels (IaaS, PaaS, SaaS) provide on-demand access to resources and offer, in particular, a high level of elasticity. With multiplying uses and a steadily rising number of users, the cloud—and the associated infrastructures and components (data center, networks, end-user devices)—is now being challenged on its environmental impacts: natural resources used, primary energy, water, CO2 emitted, etc. These environmental impacts must be measured across the entire value chain, from the construction of the building infrastructures that host IT as a Service offerings to the uses made of them, including their operation and end-of-life management.

Because while the cloud can have positive effects (resource pooling, pay-per-use activation, easier remote working…), the increase in bandwidth needs, data duplication, and the rising equipment rate required to use these services are now being questioned as part of a sustainable IT approach.

This approach needs to be made objective, based on specific functional units, to support choices and the management of these cloud services: architectures, platform types, providers, etc. This also makes it possible to compare technical solutions and highlight strategic choices, both internally and externally.

To provide fair and relevant measurements, life cycle assessment (LCA), derived from industrial methods, gives organizations a view that is both standardized[1] and realistic of the environmental impacts of their cloud services—from raw material extraction through to end-of-life recovery—including energy and water consumption, waste generation, manufacturing, distribution, and service use.

It involves measuring, for a functional unit (for example, one minute of video conferencing), the impacts of the different layers and sub-components required for the cloud service to operate: data center (building and technical infrastructure), IT infrastructure (servers, network equipment, storage…), software (including application code), user devices, and data flows transmitted over communication networks. The volume of cloud services used, as well as load and utilization rates relative to overall capacity, can be used to weight the results.

To be as realistic as possible, credible usage scenarios must be developed to assess the cloud’s environmental impacts. Thus, for a SaaS web conferencing solution provider, nine scenarios were developed by combining the connection location (EMEA, Asia, United States), the devices used, and the communication networks used (smartphone on 4G, personal computer on ADSL, multi-user conference room on ADSL).

Based on these data, organizations can produce practical reports on the cloud’s environmental impacts across the entire value chain, using indicators related to all types of potential environmental impacts: extraction or reuse of natural resources, primary energy used, water, and the amount of carbon dioxide emitted at the different stages of cloud service use.

Measuring for the sake of measuring, or a posteriori analysis, does not really make sense. The goal is to understand environmental impacts depending on how the cloud is used, so that the right decisions can be made to limit them. In this respect, the associated measurements and calculations must be automated as much as possible, for efficiency and, of course, to limit the resources devoted to them.

In terms of equipment, choices can be guided by intrinsic characteristics, but also by opportunities to optimize service life, repairability, and recycling. In the same way, it will be possible to steer energy choices (consume less, better, and more renewable energy) and architectural choices. Finally, communication aspects should not be overlooked, because it is also by mobilizing and involving users—by making what is invisible and virtual tangible—that it is possible to reduce the cloud’s environmental impact across its entire value chain.

A major challenge remains in the environmental management of the cloud: access to data within a reference framework, in order to understand the state of the art, benchmark, and of course improve. This is precisely the purpose of the NegaOctet project, led by several industry players and supported in particular by ADEME, which involves building a standard method for assessing the environmental impacts of digital services. With one main objective: to give users the means to objectively compare the environmental impacts of digital services, as is already the case, for example, with household appliances. Still in the pilot phase, NegaOctet aims to become a standard very quickly.

By Caroline Vateau, Director of the Sustainable IT Department, APL.

[1] ISO/TR 14062:2002 – Integrating environmental aspects into product design and development;
NF EN ISO 14040:2006 – Life cycle assessment: principles and framework
NF EN ISO 14044:2006 – Life cycle assessment: requirements and guidelines
NF EN ISO 14063:2010 – Environmental communication – Guidelines