Panel: Unsolved Problems in SRE
summary
Panel surveys where SRE practice actually breaks: contrasts consulting vs ownership model, warns one-size-fits-all SLOs, and argues for tool-driven production-readiness gates (validation tooling that enforces deploy policies). Covers data-reliability needs, ML services that lack clear invariants, and practical signals like capacity insights and anomaly detection used today. Debates whether predictive ML gives understanding or just black-box alerts and calls for cross-organizational research and better onboarding to reduce “keep-the-lights-on” toil.
AI-written summary · about
chapters (19)
- 0:00Introduction
- 1:13What is SRE
- 2:52Traditional SRE
- 3:44The Unsolved Problem
- 5:09Another Model
- 5:56Data is Key
- 7:02Paradigms
- 9:01Complexity
- 10:57Systems are sociotechnical
- 13:12RPCS
- 14:20Abstraction
- 17:26Ryan
- 19:32Data Science
- 21:20High Modernism
- 23:43Lack of Models
- 25:43Empiricalism
- 29:32Reliability
- 33:55Deep Conceptual Work
- 41:55Conclusion
related talks
49mWhat SRE Could Be: Systems Reliability Engineering
37mImplementing SRE in a Telco with Reliability Enhancing...
32mIs the S in SRE for “Security”?
20m5 Agile practices every SRE team should adopt
28mThe Curse of SRE Autonomy and How to Manage It
35mSRE for [cyber]security
46mThoughtful Software Engineering
28mTowards a culture where everyone is responsible for downtime