SRE 101
summary
SRE practices from Google for operating products at massive scale are presented, framing reliability in user-facing metrics (SLI/SLO/SLA) rather than server uptime and giving concrete examples (e.g., 99.999% ≈ 26 seconds downtime/month, 90% ≈ 3 days). The talk recommends aggregating logs into BigQuery for continuous observability, automating toil selectively, running blameless postmortems, monitoring carbon emissions as an ops metric, and building teams with psychological safety to speed incident response.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:45What is SRE?
- 4:34DevOps to SRE
- 8:06SRE at Google
- 10:36Key SRE components
- 13:09Automation and toil
- 14:13Availability and reliability
- 16:39Why not 100% reliability?
- 19:31SLI, SLO, SLA
- 22:05Observability in GCP
- 23:52High-performance culture
- 26:59Psychological safety & postmortems
- 28:01SRE learning resources
- 29:27Q&A
