Mark Imbriaco interviewed at Velocity 2011
summary
Heroku hosts about 150,000 apps on Amazon EC2, routinely spinning up and replacing hundreds of nodes so every instance is treated as disposable; outages are meant to be non-events rather than paged emergencies. Incident communication uses a status blog and Twitter during events and postmortems afterward, experimenting between technical timelines and prose narratives. The platform team attempts platform dogfooding with bootstrap strategies to avoid a circular dependency where the platform must run itself.
AI-written summary · about
chapters (5)
- 0:00Intro
- 0:20How many apps does Heroku host
- 1:00Whats the best mindset for handling failure
- 2:00Whats the best way to communicate after failure
- 3:00Is there such a thing as too much dog food
related talks
38mMillions of Apps: What we've Learned
25mBreaking the 2 Pizza Paradox with Platform Applications
05mIgnite - What I Learned From a Dress an On-Call Nightmare
55mDevOps @Scale (Greek Tragedy in 3 Acts)
44mIntroduction to application scaling
47mPlatform Team Structures and Concerns
31mDevOps in the Cloud: Case Studies of Amazon.com teams and their resilient architectures
43mMillions of Apps: What we’ve Learned