5 Years of Rails Scaling to 80k RPS
summary
Claim of five years to push Rails to 80k RPS across multiple datacenters with concrete wins: built an in-house load tester, introduced identity cache to cut DB load, implemented per-shop sharding and pods, added a sorting hat router for multi-DC routing, and hardened failures with toxiproxy and Semian. Result: real multi‑DC failover and 80k steady peaks; tradeoffs include a global master DB, shared worker capacity limits, and sharding-induced query restrictions.
AI-written summary · about
chapters (8)
- 0:00Start
- 7:32OPTIMIZATIONS
- 10:06IDENTITYCACHE
- 13:52OPTIMIZATION FALLACY
- 15:18SHARDING 101
- 19:17SURFACE AREA
- 24:39Resiliency Maturity Pyramid
- 26:01ACTIVEFAILOVER: 10-60s FAILOVERS
related talks
40mHow Shopify Scales Rails
27mRails Orbiting the cloud: Mastering the scalability with Kubernetes
44mIntroduction to application scaling
35mTen Years of Rails Deployment
21mTesting Rails at Scale
33mShopify in Multiple Datacenters
32mOperating Rails in Kubernetes
31mHow We Went From Two Major Outages to 99.98% Reliability in 6 Months