When Web Scale Doesn't Make Sense
summary
Argues that most apps avoid web-scale complexity by keeping data on one powerful node (Postgres) and minimizing extra stores like Redis, ElasticSearch, or RabbitMQ unless traffic or features demand them. Shows vertical scaling limits with concrete numbers — VMs with 128 cores or multi‑TB RAM, Dan Kegel’s 10,000 concurrent connections benchmark — and recommends a mono repo and using Postgres features (full‑text, JSON, LISTEN/NOTIFY) before introducing distributed services.
AI-written summary · about
chapters (6)
- 0:00Intro
- 1:30Other measures of scale
- 6:18Right abstractions
- 12:26Hard ceiling
- 16:58Atomicity
- 26:31Recommendations
related talks
48mWeb scale for the rest of us
38mThe Dark Side of Scale
47mWeb-Scale IT Engineering.
49mHypergrowth scaling made simple
01h 22mWebinar: Decomposing Applications for Deployability and Scalability
42mLarge-Scale Architecture: The Unreasonable Effectiveness of Simplicity
05mScale Fail
56mMastering Cloud Scalability and System Designs Deep dive