Managing Demand and Efficiency for Meta Multi-tenant Services
summary
Meta builds a one-stop Demand Efficiency Portal that enforces quotas and admission control, runs automatic regression detection with FB Detect, and surfaces rule-based optimization suggestions plus an Efficiency Hero tracker. The system handles thousands of daily code and experiment changes across hundreds of teams, onboarded 17 infra services and covers ~30% of Meta footprint, and yields ~5x more efficiency wins by shifting ownership attribution from endpoints to call-stack and cost-tax models and automating quota transfers and throttles.
AI-written summary · about
chapters (11)
- 0:00Demand efficiency journey
- 1:07Five-year Meta evolution
- 4:01Multi-tenant demand challenges
- 8:37Federated demand model
- 9:42Demand Efficiency Portal structure
- 11:54Product attribution for cost tracking
- 14:48Quota management mechanisms
- 15:32Admission control systems
- 16:35Rule-based optimization
- 17:37Efficiency Hero system
- 18:18Portal impact and future
related talks
50mWhen Should You Consider MetaArchitectures? Using Meta to Scale the Cloud
20mGlobal Capacity Management
25mOpening Keynote
39mPractical Container Scheduling: Juggling Optimizations, Guarantees, and Trade-Offs at Netflix
12mThe Next API Frontier: Mastering API Consumption Across Distributed Envts
44mBuilding Multi-Tenant Compute Systems in the Enterprise
30mMgmt Config: The future of your autonomous datacenter
15mEvolving Cluster Management