Introduction to application scaling
summary
Platform serves about half a million requests an hour with regular 3–5× spikes and a surprise peak of 1.2M/hr that made a customer $200,000. Practical fixes include using New Relic to find hot spots, writing an Eloquent collection extension that cut queries by ~70%, separating web/control/db roles, adding HAProxy and Redis, vertical-scaling a big DigitalOcean instance as a stopgap, then moving to Amazon Aurora for higher write throughput, and applying caching and rate limits (from ~80 to ~15 req/s) which all materially reduce outages.
AI-written summary · about
chapters (32)
- 0:00Introduction
- 2:00Jules Reference
- 2:25Agenda
- 3:20Admin Panel
- 4:00Web Stores
- 4:30Minecraft
- 5:30Traffic spikes
- 7:15Why do we need scale
- 8:00Slow performance
- 9:20Separation of concerns
- 13:10Optimization
- 14:20Dam Plus One
- 16:45Loading Relationships
- 17:50Assign Relationships to Collections
- 18:40Hardware scaling
- 19:00What needs to scale
- 19:40Vertical scaling
- 20:53Horizontal scaling
- 22:25Load balancing
- 22:55Database scaling
- 23:30Sharding
- 25:25Replication
- 25:45Masterslave
- 27:18Master replication
- 31:13Caching
- 31:30Dynamic content
- 33:33Cache
- 35:05Prewarp
- 36:41API attacks
- 37:36Rate limiting
- 38:47Cloud Flask
- 41:06Questions
related talks
47mScaling: from 0 to 25 million users
37mScaling Rails for Black Friday and Cyber Monday
24mSleeping soundly even when scaling your app to 200k request/seconds
29mEvolve your web application for scale
01h 03mThe Trail to Scale Without Fail: Rails?
41mForget Scaling: Focus on Performance
31mA Rails performance guidebook: from 0 to 1B requests/day
54mGetting to 300 million requests a month with a one man AI/Dev-Ops team