Reliability Engineering
summary
Argues that fault prevention is obsolete and teams must engineer fault tolerance instead, citing Bell-era electronic switches with a design goal of two hours downtime in 40 years (about three minutes per year), and real outages: July 8 2015 market and airline failures, Dec 2017 Atlanta power outage, June 1 2018 Visa outage affecting 52 million transactions. Recommends SRE practices: automation to eliminate toil, heavy monitoring, canary releases, failover to replicas, and an error budget tied to product velocity plus chaos engineering experiments; notes SREs earn ~50% more but face on‑call burden.
AI-written summary · about
chapters (14)
- 0:00Intro
- 3:45Number to ESS
- 7:55Did it work
- 14:20System failures
- 16:06Reliability is a system characteristic
- 19:22Site Reliability
- 20:56Internet Scale Economics
- 23:29Site Reliability Engineering
- 24:41Error Budget
- 25:33Maximum Change Velocity
- 27:50Mikey Dickerson
- 32:15Reliability Engineering
- 33:31Chaos Engineering
- 35:02Pros and Cons
