Live Q&A with Speakers – Laine Campbell
summary
Live Q&A digs into concrete tradeoffs and fixes: measured cost of site degradation/MRDR by SLAs and user engagement; anti-entropy gaps in Dynamite → Cassandra + S3 migration forced extra resiliency until complexity pushed a full migration with fallback guardrails; vacuum testing flags leaked critical dependencies; 99% of workflows finish within a week (long tail up to eight months) so a switch-over pattern handles most cases; Scribe uses multi-tenant versus isolated tiers for noisy-neighbor protection; P99 latency ~10s, five nines vs six nines debated.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 0:45How does Meta make sure that the cost of handling DRM is acceptable to the business
- 1:57How long did it take to get agreement on shared goals
- 3:11How did the lack of antientropy and dynamite affect conductor operations
- 4:42Did you expect a Tipping Point
- 5:41How do you ensure critical dependencies dont regress
- 6:52Dragons
- 10:48Keyshore
- 12:23Long tails
- 13:13What would it take to get to 6nines
- 14:54How to get to 6nines
- 20:14What happens if you go back
