Low-Carbon and Low-Latency Cloud Region Selection Project in cooperation with Green Coding Solutions Cloud services are usually deployed where infrastructure is cheap, familiar, or close to the development team. Yet the environmental impact of hosting can vary dramatically depending on where servers are located, because electricity grids differ widely in their carbon intensity. At the same time, moving a service to a lower-carbon region may not necessarily worsen the user experience: for many applications, the difference in responsiveness between several nearby or well-connected regions may be negligible. This project explores exactly this trade-off: Can we recommend cloud or hosting regions that are both low in carbon intensity and fast enough for the intended user base? The project will build on an existing prototype: carbon_map [0]. The core idea is to combine global network latency measurements with electricity grid carbon-intensity data. As a first step network data can be obtained from RIPE Atlas [1], a large-scale global measurement platform that provides distributed probes for measuring internet connectivity and ping times. Other sources could also be investigated. Carbon-intensity data can initially be based on open datasets such as the grid-intensity mappings used by the Green Web Foundation’s co2.js project [2]A more advanced version of the project could also integrate dynamic, near-real-time grid intensity data. This would allow the system not only to recommend generally low-carbon regions, but also to account for changing electricity conditions over time. In a final step, it should also be possible to include data center utilization data. Here, spot instance price can be used to estimate the utilization as described in [3]The intended outcome is a small web service that allows a service provider to specify where their main users are located. The system then recommends suitable hosting regions by balancing three objectives: Low carbon intensity of the electricity grid in the hosting region. Acceptable network responsiveness for the target user population. (Optionally) Datacenter utilization to minimize embodied emissions. For example, if most users are in Central Europe, the system might compare nearby regions and show whether a significantly lower-carbon hosting location can be chosen with little or no practical degradation in latency. This could help developers, researchers, and companies make more climate-aware infrastructure decisions without sacrificing quality of service. If time allows, the project can be extended by investigating where users are globally concentrated. By incorporating population or internet-user distribution data, the system can move beyond single-location assumptions and reason about realistic demand patterns. This raises interesting questions: Where should a service be hosted if its users are mostly in Europe? What about a globally distributed user base? How much latency penalty is introduced by choosing a cleaner grid? Are there “sweet spot” regions that provide both good connectivity and low carbon intensity? The project combines several relevant areas: sustainable software engineering, internet measurement, cloud infrastructure, data visualization, and applied web development. Depending on the student’s interests, the focus can be adjusted toward data analysis, backend development, frontend visualization, optimization, or environmental modeling. Possible tasks include: Collecting and processing latency measurements from RIPE Atlas. Mapping measurement locations and hosting regions to electricity-grid carbon intensity. Integrating static and, if possible, dynamic carbon-intensity datasets. Designing a scoring model that balances latency and carbon impact. Adding population or user-distribution data to model realistic demand. Building an interactive web interface for exploring recommended hosting regions. Evaluating example scenarios for European or global user bases. Additional connected ideas could include studying the correlation between data center locations, carbon intensity, and accessibility by leveraging public data center maps [4]. The broader vision of this project is to make carbon-aware hosting decisions easier. Instead of treating infrastructure placement as a purely economic or technical choice, this project asks how we can incorporate the climate impacts of electricity grids while still meeting users’ performance expectations. The result should be a demonstrator that shows developers where they can host software services with lower emissions and minimal compromise in responsiveness. github.com/ribalba/carbon_map atlas.ripe.net/docs/getting-started/what-is-ripe-atlas github.com/thegreenwebfoundation/co2.js/tree/main/src/data/electricity-maps Adrian Cockcroft, Don’t follow the sun: Scheduling compute workloads to chase green energy can be counter-productive, 2003 www.datacentermap.com
