All problems

IoT Telemetry

Swallow a device-fleet firehose without dropping a datapoint.

hard~22 min Pro Simulator · +140 ptsTheory +45 pts
The build

Swallow a device-fleet firehose through a reconnect storm while dashboards stay fast and the TSDB survives.

Millions of devices phone home on a schedule — until a firmware bug makes them all reconnect at once. The ingest path must buffer the firehose in a stream before it reaches the time-series store, while fleet-status dashboards read from a cache so they never depend on the hot write path.

Components in play

  • Ingest gatewayTerminates millions of device connections and splits telemetry writes from dashboard reads.
  • Telemetry streamDurable log that buffers the ping firehose so a burst never hits the store raw.
  • Ingest workersDrain the stream, batch points, and write them into the TSDB.
  • Time-series DBStores every datapoint; the one tier that must absorb the full ingest rate.
  • Fleet-status cacheServes 'is my fleet healthy' dashboard reads from memory, off the write path.

Graded on these SLOs

  • ≥ 98.00%Telemetry accepted
  • ≤ 120 msIngest-to-dashboard p99
  • ≥ 18k rpsSustain the firehose
  • ≤ 90%Ingest headroom
  • ≤ 12Lean fleet
Base 14,000 rps45s run 4 scripted failures

The brief

Design the ingest pipeline behind a fleet of IoT devices: millions of them phone home on a schedule, and a firmware bug or a returning outage can make them all reconnect at once. The ingest path has to absorb that firehose into a time-series store without losing datapoints, while fleet-status dashboards read from a cache that never depends on the write path being healthy.

Functional

  • Ingest a steady stream of device telemetry
  • Absorb a fleet-wide reconnect storm without dropping data
  • Serve 'is my fleet healthy' dashboard reads

Non-functional

  • No lost datapoints, even under a 4-5x reconnect spike
  • Ingest-to-dashboard latency stays low under load
  • Dashboard reads must not depend on the time-series DB being up

Warm-up: design decisions

Optional theory to prime the calls a senior engineer would make before you build.

Pro

Solving is a Pro feature

Unlock the on-canvas AI design coach, per-run design reviews, the chaos simulator and all 12 problems for $15/month.

  • AI design coach on your canvas — pins failure points & over-engineering
  • Honest AI design review on every run
  • All 12 problems + unlimited chaos simulator
  • Points, levels & leaderboard
Get started