Scaling Infrastructure With Confidence in the Age of AI
summary
Convex prototypes an in‑memory index cache that supports get put and invalidate, uses Shuttle randomized testing to reproduce concurrency races, and compares cached vs DB results in production until equal. The prototype yields 75% cache hit rate, cuts database CPU by about 80%, and halves latency while avoiding a new service. Tests and abstractions prioritize consistency and multi tenancy so iterations remain safe to ship.
AI-written summary · about
chapters (8)
- 0:00Introduction and background
- 1:50How Convex works
- 2:56Open Claw challenge and scaling issue
- 3:58Core requirements for cache
- 5:23Testing and production safety
- 7:07Randomized testing with Shuttle
- 8:09Key takeaways and optimizations
- 9:58Team size and hiring
related talks
21mBuilding Distributed Platforms at Extreme Scale on Fast Changing Data
19mBeyond the Node: Scaling Inference with Cluster-Wide KVCache...
16mPreview: Cloud Computing
19mBuilding a Performant Configuration-Store Layer
12mKV Cache and Long-Context Inference: Solutions for Inference at Scale
43mCaching half a billion user transactions
17mEfficient Dependency Resolution in IFC-aware Decentralized Programming
25mDistCache: Provable Load Balancing for Large-Scale Storage Systems with Distributed...