Climate compensate with a pull request

Have you ever thought about that we who work with tech can influence the climate impact of the systems we develop? This persentation is about how we can improve the carbon footprint of the services that we build.

Transcript

1. Two years ago, I received a new pair of glasses Hi! Photo by Bud Helisson on Unsplash
4. Energy consumption depends on: • Sending and receiving data • Client screen usage • Client and server CPU activity
5. The emissions from electricity varies a lot between countries
7. Emissions also come from the development process • Travel • Office • Test environments • etc Photo by Marvin Meyer on Unsplash
8. How much carbon dioxide emissions do we cause?
9. Example calculation from a project at SVT Play
10. Example calculation • Input ◦ Amount of data transferred over networks ◦ Screen/device time per type ◦ Server running time ◦ Location of clients
11. 10g carbon dioxide per hour of streaming
12. Source: max.se
14. 4 329 tonnes CO2e/year = 9 620 x ✈ Sthlm - NY Source: flysas.com
15. Then what?
16. Stop autoplaying recommendations = saved 200 tonnes of carbon dioxide per year
17. AVIF encoded images = saved 2 tonnes of carbon dioxide per year
18. What can we do to minimize the carbon footprint?
19. Reduce waisted data
20. Optimize encoding of video content 4K: 17Mb 1080p: 9Mb 720p: 4Mb
21. Use responsive images 169Kb 63Kb 37Kb 1920x1920 1024x1024 720x720 <img srcset="sprout-480w.jpg 480w, sprout-800w.jpg 800w" sizes="(max-width: 600px) 480px, 800px" src="sprout-800w.jpg" alt="Green sprout">
22. Use modern image formats such as AVIF or WebP AVIF: 50Kb WEBP: 100Kb JPG: 169Kb <picture> <source type="image/avif" srcset="sprout.avif"> <source type="image/webp" srcset="sprout.webp"> <img src="sprout.jpeg" alt="Green sprout"> </picture>
23. Use reasonable image quality levels Q=70: 169Kb Q=90: 221Kb Q=100: 478Kb
24. Source httparchive.org
25. Minimize JavaScript, avoid bloated libraries, use treeshaking etc
26. Use gzip compression for html, json, css etc
27. Check your Cache-Control headers
28. Block bots
29. Reduce waisted servers
30. Choose serverless or autoscaling if you can • With serverless, the machines can be more efficiently utilized • Autoscaling to zero means no energy consumption when it's not need
31. Stop unused test environments • Serverless • Autoscale to zero • Or schedule them to stop when not needed Photo by John Smit on Unsplash
32. Optimize CPU utilization • Performance tuning • Minimize CPU allocation • Tune garbage collection settings • etc
33. Machine learning • Choose simple models instead of deep learning when possible • Train iteratively instead of retraining from scratch periodically • Use transfer learning
34. Check your partners Check energy efficiency of your data center Check energy efficiency of any SaaS/PaaS/IaaS used Check energy efficiency of your CDN provider
35. In general, hyper scale data centers are much more efficient than small local data centers This Photo by Unknown author is licensed under CC BY.
36. Reduce waisted screen time
37. Good UX design can make a big difference
38. Design efficient workflows to minimize waisted screen time
39. Some features can reduce waist
42. Dark mode
43. Make climate impact a factor in your decisions and priorities
44. Summary 1. Many small savings can make a big difference 2. Reduce waisted data 3. Reduce waisted servers 4. Reduce waisted screen time 5. Make climate impact a factor in your decisions and priorities Photo by Bud Helisson on Unsplash
45. Attributions & sources • https://unsplash.com/photos/kqguzgvYrtM • https://unsplash.com/photos/8BNO4HUQYRw • https://unsplash.com/photos/SYTO3xs06fU • https://unsplash.com/photos/hRI8f-2WyDw • https://app.electricitymap.org/map • https://www.flysas.com/en/sustainability/emissio n-calculator/ • https://www.max.se/maten/meny/maltider/ Photo by Михаил Павленко on Unsplash