Lean Mean Green Machines - Architecting for a Greener Future - Sara Bergman - NDC Oslo 2024

Carbon efficiency in software consists of three main areas: energy efficiency (using least energy possible), hardware efficiency (minimizing embodied carbon), and carbon awareness (using more when grid is green)

Cloud providers and data centers are becoming more efficient and investing in renewable energy, but cloud users remain responsible for sustainability within their applications

Right-sizing resources through auto-scaling and proper capacity planning is crucial for both cost and carbon efficiency - idle servers waste energy even when doing nothing

The electricity grid is becoming more variable due to renewable energy adoption. Software needs to adapt through demand shaping and carbon-aware computing that shifts workloads to times/locations with cleaner energy

Storage optimization is important - implement retention policies, use appropriate storage tiers (hot/warm/cold), and avoid keeping unnecessary backup data indefinitely

Location matters - choose regions with cleaner energy grids when possible, but consider network costs and data privacy requirements when shifting workloads geographically

Building green from the start is easier than retrofitting later. Focus on architecture decisions that enable sustainability rather than premature optimization of code efficiency

The Green Software Maturity Matrix helps organizations assess their sustainability capability across different categories and progress from Level 1 (Aspiring) to Level 5 (Inspiring)

Sustainability initiatives often have co-benefits like cost savings, improved performance, and better resilience. They’re also increasingly important for employee retention

Tools and APIs are available to help measure and optimize carbon footprint, including offerings from major cloud providers and open source options like Electricity Map and Cloud Carbon Footprint