Forget Scaling: Focus on Performance
summary
Focuses on single-process speed first so scaling later is easier and shows concrete outcomes: Heroku runs about three million apps and benchmarks show 7 unicorn workers on a single Dyno and 16 unicorn workers on a double Dyno for small services like bundler API. Recommends unicorn or puma, PG stat statements for slow-query aggregation, New Relic / Librato / Skylight for visibility, CDN Sumo or CloudFront for assets, Heroku Fork for regional copies, and load tests with Blitz to validate memory and throughput.
AI-written summary · about
chapters (9)
- 0:00Start
- 1:35Ruby Task
- 5:23Single Request Server
- 6:32Concurrent Request
- 10:47Master Process
- 20:49New Relic
- 28:01Provision Follower
- 32:23CDN
- 36:43Heroku Regions
