System design interview question
Design a Distributed Time-Series Metrics Storage and Real-Time Aggregation System for Infrastructure Monitoring (Like Datadog at Scale)
Design a backend system for a global infrastructure monitoring platform that ingests, stores, and aggregates time-series metrics from millions of servers across multiple regions. The system must handle 500K metrics writes per second with sub-second query latency for dashboards and alerts, support arbitrary-dimensional aggregations (group by host, region, service, environment), and retain metrics at different granularities (1s raw, 1m rolled-up, 1h rolled-up) across 1-year retention windows. The system must guarantee metric ordering per dimension, support real-time alerting (evaluate rules within 30 seconds of metric arrival), and gracefully handle clock skew, out-of-order writes, and partial failures in a multi-region setup.
Practice this question step by step with an AI interviewer that reviews your requirements, your architecture and your trade-offs, and scores the design.