Scaling With Friends
summary
A single Rails backend supports six games and millions of daily players, with concrete ops: about 400+ app servers running ~35 Unicorns each, 125+ MySQL shards (each with a slave), ~70 memcache servers, Redis for id generation and Resque for background work. Real problems and fixes include the move‑pocalypse and chat id overflow leading to bigint rollovers and table partitioning then sharding, cache strategy changes, staged rollouts, added tests, and a shift from XML to JSON endpoints — all incremental fixes that actually work.
AI-written summary · about
chapters (23)
- 0:00Intro
- 1:14History
- 2:06Three Years
- 3:09Ecosystem
- 4:22Alec Baldwin
- 5:50Rails cant scale
- 6:17Rails is sexy
- 7:07Backwards compatibility
- 10:58What did we do
- 15:28Pains
- 17:23The Problem
- 18:30Movepocalypse
- 21:35Analysis
- 24:03How did we deal with the movepocalypse
- 28:11Sharding
- 31:27Cache
- 34:59Caching
- 36:03Scale
- 38:36Improvement
- 43:55Monitoring
- 44:52Hiring
- 45:18Words with Friends
- 45:40Outro
related talks
56mScaling Rails API to Write-Heavy Traffic
29mScaling Rails to 2M MySQL requests per second
46mRuby & Erlang: at Scale with Style
37mScaling Rails for Black Friday and Cyber Monday
34mShared Nothing Architecture - How to Scale to Infinity
47mScaling: from 0 to 25 million users
01h 03mRuby at Scale
40mHow Shopify Scales Rails