Carbon accounting in the Cloud: a methodology for allocating emissions across data center users

TL;DR

This paper tackles the problem of fairly allocating electricity use and carbon emissions in Google-scale data centers across thousands of internal users and services. It introduces a mixed physical/economic allocation framework aligned with the Greenhouse Gas Protocol Scope 3, separating idle and dynamic power and reallocating energy from internal shared services to end users, then attributing emissions to Google Cloud customers using region-specific carbon intensity data. Key contributions include a complete methodology with clear boundaries, detailed formulas for idle/dynamic power, multi-round reallocations for shared services, and SKU- and region-level emission allocations that tie to revenue data and pricing. The approach enables granular, location-based carbon reporting for enterprise customers and supports market-based reporting and optimization of data-center operations, while remaining scalable for Google’s global infrastructure. This work has practical impact by providing actionable, product-level carbon-footprint data that informs reductions in energy use and emissions across cloud services and regions, and it lays the groundwork for incorporating market-based accounting and manufacturing-emissions considerations in future work.

Abstract

This paper presents a methodology for allocating energy consumption to multiple users of shared data center machines, infrastructure, and software. Google uses this methodology to provide carbon reporting data for enterprise customers of multiple Google products, including Google Cloud and Workspace. The approach documented here advances the state-of-the-art of large scale Cloud carbon reporting systems. It uses detailed, granular measurement data on machine energy consumption. In addition, it uses physical factors for allocating energy consumption and carbon emissions--preferred by the Greenhouse Gas Protocol's Scope 3 Reporting Standard. Specifically, the approach described here allocates machine energy consumption based on a combination of data center resource reservations and hourly measured resource usage. It also accounts for Google's own internal use of shared software services, reallocating energy use to the users of those shared services. Finally, it uses hourly, location-specific estimates of carbon intensity to precisely measure carbon emissions of users in a global fleet of data centers.

Figures (7)

• Figure 1: A simplified example of energy allocation in a datacenter with diurnal usage patterns and no shared services. One user (blue) is responsible for all production jobs and resource allocations; the other (orange) runs non-production jobs. In the upper chart, the example utilization values would be measured. In the lower chart, power is measured, while the proportion of power to allocate to each user is calculated based on utilization measurements and the described formulas.
• Figure 2: The amount of idle power, dynamic power, and total power consumed per hour for a specific Google Cloud cluster, as a percentage of the weekly average total power.
• Figure 3: High-level overview of the process of allocating energy and carbon emissions to internal Google users.
• Figure 4: Example of net cost flows in the resource economy used to allocate carbon of internal Google shared services.
• Figure 5: Sankey diagram showing illustrative energy flows allocated to Cloud and internal Google users of Google Cloud. This represents, using an illustrative example, a portion of the data center energy consumption at Google that pertains directly to Google Cloud.