Usage Estimation Methods

This section documents how the Ada Carbon Monitoring system estimates electricity usage and carbon footprint from CPU metrics.

Overview

The carbon calculation pipeline:

CPU Metrics (Prometheus)
        │
        ▼
┌───────────────────────┐
│  Electricity          │
│  Estimation           │
│  (CPU → kWh)          │
└───────────────────────┘
        │
        ▼
┌───────────────────────┐
│  Carbon Intensity     │
│  (UK Grid API)        │
│  (gCO2/kWh)           │
└───────────────────────┘
        │
        ▼
┌───────────────────────┐
│  Carbon Footprint     │
│  (kWh × Intensity)    │
│  (gCO2eq)             │
└───────────────────────┘
        │
        ▼
┌───────────────────────┐
│  Attribution          │
│  (User / Group)       │
└───────────────────────┘

Key Concepts

Busy vs Idle

CPU time is categorized as:

Busy: Active computation (user, system, nice, irq, softirq, steal)
Idle: Waiting for work (idle, iowait)

Power Model

Different power consumption for each state:

Busy power: 12W per CPU core
Idle power: 1W per CPU core

Carbon Intensity

UK grid carbon intensity varies throughout the day:

Low: 50-100 gCO2/kWh (high renewable generation)
Moderate: 100-200 gCO2/kWh (mixed generation)
High: 200-300+ gCO2/kWh (high fossil fuel generation)

Calculation Steps

Step 1: Get CPU Metrics

Query Prometheus for CPU seconds:

cpu_data = prometheus_client.get_cpu_usage_breakdown(
    timestamp=datetime.now(),
    cloud_project_name="IDAaaS",
    host="172.16.100.50"
)
# Returns: {"busy": 1234.5, "idle": 56789.0}

Step 2: Estimate Electricity

Convert CPU seconds to kWh:

energy_kwh = electricity_estimator.estimate_total_kwh(
    busy_seconds=cpu_data["busy"],
    idle_seconds=cpu_data["idle"]
)
# Formula: (12W × busy + 1W × idle) / 3,600,000

Step 3: Get Carbon Intensity

Query UK Grid API:

intensity = carbon_client.get_current_intensity()
# Returns: {"intensity": 185, "index": "moderate"}

Step 4: Calculate Carbon

Multiply energy by intensity:

carbon_gco2eq = energy_kwh * intensity["intensity"]

Step 5: Attribute to User/Group

Match to workspace owner or facility:

user = mongo_client.get_user_by_platform_name(workspace["owner"])
# Or for groups:
group = f"{cloud_project}_{machine_name}"

Pages in This Section

Page	Description
Electricity	Estimating kWh from CPU seconds
Carbon Footprint	Calculating gCO2eq from kWh
User Attribution	Attributing carbon to users
Group Attribution	Attributing carbon to groups

Implementation

The calculations are implemented in ada-carbon-monitoring-api:

Module	Purpose
`src/calculators/electricity_estimator.py`	CPU → kWh conversion
`src/calculators/carbon_calculator.py`	kWh → gCO2eq conversion
`src/calculators/carbon_equivalency.py`	gCO2eq → equivalencies
`src/clients/prometheus_client.py`	CPU metric queries
`src/clients/carbon_intensity_client.py`	UK Grid API client
`src/models/workspace_tracker.py`	Orchestration and attribution

Example: Full Calculation

from datetime import datetime
from src.models.workspace_tracker import WorkspaceTracker

# Initialize tracker
tracker = WorkspaceTracker(
    prometheus_url="https://prometheus.example.com",
    mongo_uri="mongodb://localhost:27017"
)

# Track a group
entry = tracker.track_group(
    cloud_project_name="IDAaaS",
    machine_name="Muon",
    timestamp=datetime.now()
)

print(f"CPU Time: {entry.busy_cpu_seconds_total + entry.idle_cpu_seconds_total:.1f} seconds")
print(f"Energy: {entry.total_usage_kwh:.4f} kWh")
print(f"Carbon: {entry.total_usage_gco2eq:.2f} gCO2eq")
print(f"Intensity: {entry.carbon_intensity_g_per_kwh:.0f} gCO2/kWh")

Output:

CPU Time: 287776.7 seconds
Energy: 0.0816 kWh
Carbon: 18.89 gCO2eq
Intensity: 232 gCO2/kWh