Carbon Estimation Plugin

Overview

The carbon plugin estimates operational emissions for Amazon EC2 t3, m5, c5, and r5 instances using a pinned Cloud Carbon Footprint dataset. It works offline without AWS credentials. Core adds carbon to the monetary result while retaining its amount, currency, source, and resource count.

Install and Run

From the FinFocus repository root:

make install-carbon
finfocus plugin list
finfocus cost projected --pulumi-json plan.json

The plugin releases independently under carbon-vX.Y.Z; catalog installation requires the first release’s downloadable assets. Until then, build from source. Install a monetary pricing plugin or configure local pricing specifications for dollar estimates. A carbon-only installation still reports NO_COST_DATA for missing monetary pricing.

Configure Utilization

FINFOCUS_CARBON_UTILIZATION=0.25 finfocus cost projected --pulumi-json plan.json

The default is 0.5. Values must be finite fractions in [0, 1]. The plugin advertises utilization in its settings metadata too. Zero utilization retains idle compute power, and provisioned memory is not scaled by CPU utilization.

Interpret Results

The canonical carbon_footprint metric contains a monthly total in gCO2e, based on 730 hours. At the default utilization, t3.medium in us-east-1 estimates 1437.00152 grams CO2e/month. Existing output converts qualifying totals into carbon equivalencies. A monetary source’s existing carbon takes precedence; carbon failures do not discard successful monetary prices.

Carbon enrichment runs again on monetary cache hits, so changed assumptions take effect on the next query. It never turns missing pricing into a free-resource claim or duplicates resource rows.

Coverage and Limitations

There are 34 supported sizes and 38 regions in the pinned snapshot. Unknown sizes/regions are declined explicitly. These are operational compute and memory estimates, not measurements or total lifecycle emissions. They exclude storage, networking, embodied emissions, actual billing data, other clouds, and Kubernetes workload allocation.

See the plugin README for exact supported sizes, installation paths, and development commands, and data provenance for the formula, unit conventions, coefficient revision, and source attribution. The underlying approach is described by Cloud Carbon Footprint.