ServiceRouter: Hyperscale and Minimal Cost Service Mesh
summary
ServiceRouter runs Meta’s global service mesh with millions of L7 routers and routes tens of billions of RPCs per second while keeping the main data plane in-process: 99% via embedded library, not sidecars, cutting hardware needs by orders of magnitude. RIB metadata totals ~12 GB (60 TB replicated), write rate ~335/s. It natively supports sharded services (~68% of RPCs), uses locality rings for latency-aware sampling, and XRS autonomously rebalances (example: shifted 5% then 25% of traffic to relieve overloads).
AI-written summary · about
chapters (4)
related talks
21mServiceRouter: Hyperscale Service Mesh
47mMonolith to Microserv. at 20M+ requests per second: Latency and Scale Challenges
28mHow a Service Mesh Helped Us Build Production Cloud-Native Services
35mNavigating the Service Mesh Ecosystem
38mRouting Your Microservices Architecture
39mGrow with Less Pains - Meshing From Monolith to Microservices
34mDecentralized Routing For a Sharded Application On Service Mesh
29mYou Probably DON’T Need a Service Mesh!