Perfect Scalability
summary
Defines scalability vs performance, claims perfect scalability means near-linear capacity as nodes are added, and shows why common walls appear (load balancer hotspot in a simple add-two-integers example, ordering, shared state, coherency delays). Recommends denormalize and give each service private data, avoid no shared database, use command sourcing (persist commands first) and idempotency, treat communication as cost, use Kafka as shock absorber, and build monitoring plus rolling-upgrade DevOps.
AI-written summary · about
chapters (37)
- 0:00Intro
- 0:14What?
- 6:23A Perfectly Scalable System
- 7:00Why does this system scale?
- 8:42How do we fix it?
- 9:35Closer to Perfection
- 10:32Amdahl's Law
- 11:41Gunther's Law
- 15:21Avoiding the Wall
- 15:51The Enemies of Scalability
- 16:47Avoiding Contention
- 17:36Avoiding Sharing
- 18:18Avoiding Shared State
- 19:40Limit Communication
- 20:54Avoiding Ordering
- 21:54Avoiding System Failure
- 22:41Avoid Linear Time
- 24:35Designing for Perfect Scalability
- 25:56Scalable Architecture
- 26:44Scaling Down
- 27:54Ultra Large Scale Systems
- 30:33Spike Load
- 31:25Command-Sourcing
- 32:41Degrading Gracefully
- 33:50Microservices
- 35:40Simple is Good
- 36:58Persistence is (Not) Futile
- 38:16Keep your hands out of "OPD"
- 39:01Distributed Transactions
- 39:36Allow me to repeat myself..
- 40:29DDD and Scalability Onion Architecture
- 41:28Balancing Scalability
- 42:22Trust, but Verify
- 42:54The Log is not Enough
- 43:49DevOps Matters
- 44:54Conclusion
- 45:51Questions?
related talks
32mScaling Done Right
38mThe Dark Side of Scale
44mConcurrency + Distribution = Scalability + Availability(...)
18mScaling APIs to Billions: Extreme-Scale Reliability
27mHow can you Scale if you Don’t Trust It?
31mMastering the Balancing Act: Scaling Infrastructure for Growth and Stability
45mPerformance-by-Design: Embedding Intelligent Scaling and Guardrails Into Platform Engineering
01h 02mTools for simplified large scale systems