Paid to Sleep
summary
Interrupted sleep shows cognitive decline after about three years and irreversible damage after ten years, so on-call needs redesign not platitudes. Recommendations: pay a retainer for being available; build reliable systems with retries, failovers and documentation that passes the 3am test; run disaster drills (Google-style “dirt week”); tune alerts (only notify when meaningful, fix an alert that fired at 6am for a file not due until 8am); use multi-discipline rotas and a blame-free culture to avoid fatal mistakes like pkill -9 on a live MySQL.
AI-written summary · about
chapters (17)
- 0:00Introduction to paid sleep
- 1:08Science of interrupted sleep
- 3:16On-call in the gig economy
- 5:43Effects of sleep deprivation
- 7:51Speaker’s on-call experience
- 9:56Three keys to paid sleep
- 10:18Reliable systems overview
- 11:23Documentation and training
- 12:48Incident response drills
- 14:57Sensible alerting and monitoring
- 17:24Multi-discipline on-call rotors
- 19:50Culture of learning and support
- 21:36Relying on service desks
- 21:56Tuning alert thresholds
- 23:21Human factors in on-call
- 24:43Control what you can
- 25:07Conclusion and call to action
related talks
36mGetting back to sleep as soon as possible for the on-call developer
41mSleeping with one eye open: production support
06mIgnite: Key Factors for Successful 24×7 On-Call Duty
22mHumane On-call
05mOn-Call in a World Full of Toil
19mKeynote: Shut Your Laptop and Go to Sleep!
28mBuilding a healthy on-call culture
49mAgainst On-Call: A Polemic