End-to-End Design

Design a metrics monitoring system. Millions of hosts and services emit numeric measurements — request rates, latencies, queue depth, error counts — and engineers watch them on dashboards and get paged when something breaks.

In scope: ingesting metric points from many sources, storing them as queryable time-series, answering range queries for dashboards, and evaluating alert rules that page someone. Out of scope: log aggregation and distributed tracing (different systems with different storage models), the dashboard visualization frontend, and machine-learning anomaly detection.