Design a Metrics Storage Platform

Design a distributed backend for storing and querying application metrics. The system should collect time-series metrics from many services running across multiple machines and regions. Each metric sample contains: - metric name - labels or tags such as service, host, region, and environment - timestamp - numeric value Requirements: - high write throughput for continuous metric ingestion - support counters, gauges, and histogram-like metrics - allow users to query by metric name, label filters, and time range - power dashboards and alerting with low latency for recent data - keep recent data fast to query while retaining historical data cheaply - scale horizontally as the number of services and series grows - tolerate machine and zone failures - support retention, compaction, and downsampling - optionally support multi-tenant isolation and rate limits Describe the API, storage model, sharding strategy, ingestion path, query path, failure handling, and major trade-offs.