Metrics & Monitoring
Never drop a data point — and keep dashboards fast while the firehose spikes.
Keep ingesting the metrics firehose and serving dashboards while a cardinality spike triples volume and a writer dies.
Ingest is a write firehose that must never be dropped, while dashboards need hot reads served fast. A stream buffers the firehose ahead of the writer fleet so a burst doesn't overwhelm the time-series store, and a dashboard cache keeps queries off the same store entirely.
Components in play
- Ingest gateway — Accepts agent pushes and routes dashboard queries.
- Metrics stream — Buffers the write firehose ahead of the writer fleet.
- TSDB writers — Drains the stream and persists points to the store.
- Dashboard cache — Serves hot dashboard queries from memory.
- Time-series store — Durable metric history behind ingest and dashboards.
Graded on these SLOs
- ≥ 97.00%Points ingested & queries served
- ≤ 250 msDashboard query p99
- ≥ 8.0k rpsSustain the firehose
- ≤ 88%Writer headroom
- ≤ 10Lean footprint
The brief
Design a metrics ingest and dashboard platform, like a lightweight Datadog. Agents on every host push metric points continuously; engineers view dashboards built from that same data. Ingest is a write firehose that must never silently drop points, and it gets worse exactly when it matters most — during an incident, every service adds debug tags and cardinality spikes.
Functional
- Accept metric points from agents at high, continuous volume
- Serve dashboard queries over recent and historical data
- Retain enough history for on-call to review trends
Non-functional
- Ingest is write-heavy and bursty — it must never silently drop points
- Dashboard reads must stay fast even during an ingest spike
- Ingest and dashboard reads shouldn't contend for the same resources
Warm-up: design decisions
Optional theory to prime the calls a senior engineer would make before you build.
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