Lightning Talks
summary
Nine four-minute briefs give concrete takeaways: an argument for measuring the meantime to recovery of engineers and scheduling to prevent burnout; GCE rollout uses canary equivalence testing and pre-aggregated VM diagnostics to stop regressions; a multicloud MariaDB cluster migration required repeatable runbooks and two switchovers; Workspace AI launches expose LLMs as an expensive, slow dependency — treat them like a LLM cost-as-database with caching and abuse protection; a one-to-one Envoy in Rust replacement shows TLS and perf wins; SRE cost, SLO, and metric patterns finish out the set.
AI-written summary · about
chapters (12)
- 0:00Lightning talks intro
- 0:17Mean time to recovery
- 3:30Building resilient teams
- 3:51Reliability monitoring with canaries
- 7:23Cloud migration story
- 10:44LLM products in 67 days
- 15:00Rewriting Envoy in Rust
- 17:51Kubernetes cost strategies
- 22:04SRE beyond tech
- 25:57Rethinking service level indicators
- 30:11NASA Voyager lessons
- 34:04Closing remarks & Q&A
