Surviving your First Big Customer as a Startup
summary
Duffel scales a REST JSON API that searches and books flights where single searches parse megabytes of XML and a booking can require hundreds of searches per booking. The team builds a separate production-like environment, runs structured load tests (they even wrote a custom load tester but caution against it), adds extensive metrics/tracing and defines SLIs/SLOs, enforces concurrency and rate limits with Regulator, and creates a repeatable recipe so a one-week rush by a big customer doesn’t crash the system.
AI-written summary · about
chapters (19)
- 0:00Start
- 0:54The Excitement of the First Big Customer
- 2:56Procrastination
- 6:23The Goal of every Startup Is To Grow
- 13:28Can Our System Actually Handle this Traffic
- 14:14Load Testing
- 17:37Load Testing Client
- 19:46To Calculate the Capacity of Your System
- 19:58Critical Endpoint
- 21:30The Issues That You Might Come across while Load Testing
- 23:44Structured Load Testing
- 27:15Scaling Safely
- 27:49Service Level Indicators
- 33:31Concurrency
- 38:52Rate Limits
- 40:27Observability
- 42:41When Doing the Load Test How Did You Manage the Third Party Api Request Did You Mock All those Requests
- 46:16Did It Match the Capacity of the Production System
- 52:47How Did You Set Limits to the System Did You Use a Library like a Regulator or Did You Develop
related talks
45mI set fire to millions of Expedia's dollars and all I got was this talk
24mHow to design your team for scale
18mMaking what you have enough
28mFrom Startup to Scale Up: How to Scale Engineering teams
30mArchitecture & Teams at Less-than-Google Scale
50mScaling for Global Growth with Modern Cloud
31mManaging Growing Pains: Thinking Big While Being Small
38mFrom scratch to production in three months: the AI-first delivery model