Path to production: An environmentally sustainable approach
summary
Training one ML model emits as much CO2 as 125 roundtrip flights New York–Beijing, the internet uses ~10% of global electricity and tech accounts for ~3–3.6% of emissions. Concrete levers include reducing network chatter via consolidation, using serverless and autoscaling, optimizing databases with indexes and caching, compressing transit, shifting workloads to regions with cleaner grids, and tracking progress with Cloud Carbon Footprint or a Software Carbon Intensity metric. Ecosia and Holaluz show measurable wins (Holaluz cut infra costs ~3%).
AI-written summary · about
chapters (19)
- 0:00Start
- 4:27Green Technology Challenge
- 6:28How we think about green software
- 8:42Microservices optimisation
- 11:57Performance and Latency
- 12:49Reduce Data storage Adopt edge computing
- 14:12Infrastructure Sprawl
- 14:47Do you need that test environment?
- 16:01Optimising network traffic
- 17:13Operationalise green software
- 19:10Hardware Procurement
- 21:34AWS Sustainability commitment
- 22:13Temporal Efficiency
- 23:12The Software Carbon Intensity (SCI)
- 27:07Ecosia Search Engine
- 29:49Platform teams
- 30:44Benefits to technologists
- 32:11Green sensible defaults
- 33:16Green software - Tooling landscape
related talks
41mWriting Greener Software Even When You Are Stuck On-Prem
25mBuilding a Sustainable Future
26mWhy sustainability makes sense for tech
01h 02mJoint Keynote Address: Scaling AI Sustainably: An Uncharted Territory
20mBuilding GenAI That Doesn't Cost the Earth
57mThe Next Decade of Software Is About Climate - What Is the Role of ML?
46mMaking Software & Data Architectures Sustainable
49mArchitecting for Sustainability: A Software Architects Guide to Green Software Keynote