AWS Well-Architected Sustainability pillar design principles

• 1. Understand your impact — You cannot optimize what you cannot measure. AWS Customer Carbon Footprint Tool (in Cost Explorer) shows your estimated carbon emissions by service and region. Enable it and establish a baseline before optimizing.
• 2. Establish sustainability goals — Set measurable targets: "Reduce carbon emissions 20% by Q4" or "Reduce compute cost per active user 30%." Tie sustainability goals to efficiency metrics (cost per user, energy per transaction) to make them actionable.
• 3. Maximize utilization — The most sustainable instance is one running at 80%+ utilization, not 10%. Right-size aggressively. Use autoscaling. Consolidate low-utilization workloads onto fewer, larger instances (better efficiency than many small instances). Bin-pack containers with Kubernetes resource requests and limits.
• 4. Use managed services — Managed services (RDS, DynamoDB, EKS, Lambda) share underlying hardware across thousands of customers with sophisticated scheduling. AWS's utilization of their fleet is far higher than any single customer could achieve independently. Using managed services is inherently more sustainable than self-managed.
• 5. Use efficient hardware — AWS Graviton (ARM-based) processors deliver up to 60% better energy efficiency per transaction versus comparable x86 instances. Many workloads (web servers, containerized microservices, data processing) run 20–40% cheaper and with lower carbon on Graviton. Evaluate for your workload.
• 6. Reduce downstream impact — The carbon from client devices (mobile phones, laptops, browsers) consuming your service counts. Smaller API payloads (JSON compression, pagination, field selection via GraphQL), optimized images (WebP, lazy loading, CDN), and efficient frontends (smaller bundles, fewer requests) all reduce client-side energy consumption.

Changes with dual benefit: lower carbon AND lower cost

• Auto-shutdown non-production environments — Dev, staging, and QA environments running 24/7 waste 65–75% of their capacity (nobody works nights and weekends). Lambda or EventBridge rule to stop EC2/RDS instances at 7 PM, start at 8 AM. Savings: 60-75% on non-prod compute.
• Migrate to Graviton — AWS Graviton3 is up to 60% more energy-efficient and 20-40% cheaper than comparable x86. Most containerized workloads run unchanged on ARM. Start with new services; migrate existing ones during instance refresh.
• S3 storage class optimization — Move objects not accessed in 90 days to S3 Glacier Instant Retrieval. S3 Intelligent-Tiering automates this. Cold storage uses less active hardware = less energy. Typical savings: 70% on storage for compliance/audit logs.
• Consolidate idle RDS instances — Every team running their own RDS t3.small for dev means 20 databases at 5% utilization. Replace with one RDS instance or Aurora Serverless that scales to zero between dev sessions.
• Choose a greener region — AWS publishes carbon intensity by region. us-west-2 (Oregon) and eu-west-1 (Ireland) run on high renewable energy percentages. eu-central-1 (Frankfurt) is coal-heavy. If you have flexibility, region choice affects carbon footprint by 3–10×.