Resilient
summary
flipkart handles peaks where a single minute of downtime costs about $2,000 and peak minutes can cost ~$8,000, so designs favor fail fast timeouts, bounded queues and memory/CPU limits. an ebook checkout prototype shows an external payment timeout can pile messages into a queue; adding a circuit breaker plus fallback queue keeps response times low and new orders meeting SLA. practical fixes include nonblocking UDP sockets, worker monitors that restart high‑memory processes, and libraries like semian by shopify or Hystrix for JRuby.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:11Why we care about resilience
- 6:04Resilience from the start
- 6:35Disclaimer
- 8:29Patterns
- 9:15Queueing Theory
- 10:28Utilization
- 12:10Example
- 14:09Circuit Breaker
- 15:29Bounding
- 15:56Timeouts
- 22:58Circuit Breakers
- 24:57Bulkhead
- 28:17Steady State
- 29:13Log Rotation
- 29:48Archiving
related talks
42mHow to stay alive even when others go down:...
29mBlack Friday - Lessons in Resiliency and Incident Response at Shopify
29mResilient by Design
34mBuilding High Resiliency into Digital Solutions
40mZero-downtime payment platforms
32mSurviving Black Friday - A resilience engineering tale
26mThe tools and techniques behind resilient systems
47mUptime == Money: High Availability