Maximal Fun Times: Monitoring Theme Parks (and distributed systems)
summary
Compares theme-park metrics to distributed systems and shows specific levers: replace synchronous email/PDF/payment calls with a persistent queue (MSMQ, Rabbit, Azure/AWS functions), measure per-component maximum throughput from seat count, ride duration and saturation, and compute estimated queue wait time to find least responsive services. Builds queuing, measures message duration, concurrency and backlog, then raises concurrency or reduces processing time; discovers scaling hits (DB deadlocks, rate limits) and fixes bottlenecks, which cuts user complaints and latency.
AI-written summary · about
chapters (21)
- 0:00Intro
- 0:19Who is Mike
- 0:45Mikes first system
- 3:46Disk space monitoring
- 8:00Microservices
- 10:05Disneyland
- 12:21What is throughput
- 14:41Flatlines
- 16:02Roller coaster example
- 17:33Maximum throughput formula
- 19:08Experience concurrency
- 20:04Experience throughput
- 20:56Saturation
- 22:36Saturation Graph
- 23:31Full Saturation
- 25:30Queue Length
- 27:47Key Wait Time
- 30:18Queue Wait Time
- 33:49Back to work
- 37:37Increasing concurrency
- 41:45Wrapup
related talks
37mSick of Rings: The Great Battle of Monitoring Systems, Part I
34mThe Holy Grail of Systems Analysis: from What to Where to Why
19mThe Data of Infrastructures
30mRails Performance monitoring: A primer for developers
18mJon Schneider at SpringOne Platform 2017
38mLogging, Instrumentation, Dashboards, Alerts
38mEffective App Monitoring in Production
38mTechniques and Tools for a Coherent Discussion about Performance in Complex...