Best Practice: Carbon-Aware Region- und Placement-Strategie

# providers.tf – Green Region als Standard-Konfiguration
provider "aws" {
  region = "eu-north-1"  # Stockholm – ~95% renewable energy
  # Sustainability Note: eu-north-1 selected per WAF-SUS-030
  # Carbon intensity: ~10 gCO2eq/kWh vs. us-east-1 ~300 gCO2eq/kWh
}

# Für Multi-Region: Green als Default, andere nur mit Begründung
provider "aws" {
  alias  = "primary"
  region = "eu-north-1"  # Green primary
}

provider "aws" {
  alias  = "dr"
  region = "eu-west-1"  # Green DR (Ireland, ~80% renewable)
}

# Variable für Region mit erlaubten Green-Regionen
variable "aws_region" {
  type        = string
  description = "AWS region for deployment. Must be a pre-approved green region."
  default     = "eu-north-1"

  validation {
    condition = contains([
      "eu-north-1",   # Stockholm (95% renewable)
      "eu-west-1",    # Dublin (80% renewable)
      "us-west-2",    # Oregon (89% renewable)
      "ca-central-1", # Montreal (85% renewable)
      "eu-central-1", # Frankfurt (60% renewable – acceptable with justification)
    ], var.aws_region)
    error_message = "Region must be a pre-approved green region. For high-carbon regions, document justification in ADR."
  }
}

# carbon_aware_router.py
# Wählt die Region mit niedrigster Carbon Intensity für Batch-Jobs

import requests
import json
from typing import Optional

class CarbonAwareRouter:
    """
    Wählt die optimale Region basierend auf aktueller Carbon Intensity
    Nutzt electricityMaps API für Echtzeit-Daten
    """

    REGION_TO_GRID = {
        "eu-north-1": "SE",     # Schweden
        "eu-west-1":  "IE",     # Irland
        "eu-central-1": "DE",   # Deutschland
        "us-west-2":  "US-NW",  # US Northwest
        "us-east-1":  "US-MIDA" # US Mid-Atlantic
    }

    def __init__(self, api_key: str):
        self.api_key = api_key
        self.base_url = "https://api.electricitymap.org/v3"

    def get_carbon_intensity(self, grid_zone: str) -> float:
        """Gibt Carbon Intensity in gCO2eq/kWh zurück"""
        response = requests.get(
            f"{self.base_url}/carbon-intensity/latest",
            params={"zone": grid_zone},
            headers={"auth-token": self.api_key}
        )
        data = response.json()
        return data.get("carbonIntensity", 999)

    def select_greenest_region(self, candidate_regions: list[str]) -> tuple[str, float]:
        """Wählt die Region mit niedrigster Carbon Intensity"""
        intensities = {}

        for region in candidate_regions:
            grid = self.REGION_TO_GRID.get(region)
            if grid:
                intensity = self.get_carbon_intensity(grid)
                intensities[region] = intensity
                print(f"  {region}: {intensity:.0f} gCO2eq/kWh")

        if intensities:
            best_region = min(intensities, key=intensities.get)
            return best_region, intensities[best_region]
        return candidate_regions[0], 999

# Verwendung:
router = CarbonAwareRouter(api_key="your-electricity-maps-key")
best_region, intensity = router.select_greenest_region([
    "eu-north-1", "eu-west-1", "eu-central-1"
])
print(f"Optimal region: {best_region} ({intensity:.0f} gCO2eq/kWh)")

# ADR-042: Deployment Region für Analytics Pipeline

**Status**: Accepted
**Datum**: 2026-01-15
**Entscheider**: Platform Architecture Board

## Kontext

Die neue Analytics Pipeline verarbeitet Batch-Daten ohne Latenz-SLA.
Keine Data-Residency-Anforderungen.

## Entscheidung

**Primäre Region**: eu-north-1 (Stockholm)

## Carbon Impact Assessment (WAF-SUS-030)

| Region         | CO₂ Intensität | Erneuerbare | Begründung              |
|----------------|---------------|-------------|-------------------------|
| eu-north-1     | ~10 gCO₂/kWh  | 95%         | ✅ Ausgewählt – grünste  |
| us-west-2      | ~25 gCO₂/kWh  | 89%         | Alternative (Fallback)  |
| us-east-1      | ~300 gCO₂/kWh | 40%         | ❌ Abgelehnt – Carbon    |

**Geschätzte Emissionsreduktion**: ~85% vs. us-east-1 (230 tCO₂e/Jahr gespart)

## Konsequenzen

- Latenz für europäische Nutzer optimal
- Jährliche Carbon-Einsparung: ~230 tCO₂e gegenüber us-east-1