Reliability A to Z
summary
Meta describes building a new data center design with liquid cooling to support AI racks that are >4x denser, plus a pivot to heterogeneous GPU generations and 100x increase in job/container scale. Change management standardizes rollouts via ICSP paced rollouts and Conveyor; model serving reuses existing infra and a surprise: Threads hits 100 million signups that rely on the SEV tool for real-time coordination, and monitoring plus blameless postmortems drive fixes.
AI-written summary · about
chapters (6)
- 0:00Introduction and panel
- 2:40Establishing baselines
- 8:07Growth and resilience
- 15:01Organizational agility
- 17:15Reliability in the AI era
- 25:12Culture and future outlook
related talks
55mA Decade of Progress: The Evolution of Production Engineering and Reliability
28mReliability Engineering: What we learnt so far
43mSRE From Inception to Global Scale
12mDigital & Data/AI Strategy and Transformation
26mScaling engineering organizations
18mAchieving Innovation and Stability for Enterprises
43mThe SRE I Aspire to Be
52mThe Only Constant Is Change: Lessons from a 25 Year SRE Career