Scaling Rails for Black Friday and Cyber Monday
summary
Platform serves 150,000 shops and peaks near a million requests per minute; Black Friday/Cyber Monday hits ~600k req/min and ~6,000 orders/min. Scaling choices are concrete: a cashable gem implements generational page caching with ~60% hit rates to avoid DB and Liquid rendering, identity cache marshals ActiveRecord into memcached to cut MySQL queries, backgrounding and throttled Resque jobs move slow work out of Unicorn, toxiproxy and tests force resiliency, and a senior circuit breaker fails shards fast so the site keeps serving.
AI-written summary · about
chapters (11)
- 0:00Start
- 5:29generational caching
- 9:50full model caching
- 13:59throttling
- 16:55load testing
- 17:08genghis khan
- 20:28resiliency
- 21:49minor dependencies take you down
- 26:42how can we fail fast?
- 28:39what else can go wrong?
- 29:14manual circuit breakers around external dependencies
related talks
41mHow Shopify Scales Rails
40mHow Shopify Scales Rails
28mThe Real Black Friday aka How To Scale an Unscalable Service
23mHow Shopify Sharded Rails
35mSurviving Black Friday: Tales from an e-Commerce Engineer
31mA Rails performance guidebook: from 0 to 1B requests/day
44mIntroduction to application scaling
46mScaling With Friends