Bye Bye Timestream: How We Migrated 5.9 Billion Metrics a Day to ClickHouse, With Zero Downtime, 85% Lower Cost and 20x performance improvement

Komodor migrated its metrics platform from Amazon Timestream to ClickHouse Cloud after AWS restricted new access to Timestream LiveAnalytics and the company encountered high, unpredictable costs, limited data mutability, aggregation constraints, weak observability, and no local testing environment. The seven-month project moved approximately 5.9 billion metric records per day through a feature-flagged, reversible process involving dual writes, shadow queries, parity validation, phased account cutovers, and eventual Timestream decommissioning, allowing production traffic to continue without customer-facing downtime. After testing InfluxDB 3, TimescaleDB, and ClickHouse under production-scale traffic, the team selected ClickHouse for its ingestion capacity, SQL support, low latency, compression, materialized views, Docker-based local development, and projected monthly cost of $5,000–$6,000 versus about $38,000 for Timestream. Extensive unit, component, parity, and live shadow testing, supplemented by AI-assisted SQL generation and validation reporting, revealed unexpected data discrepancies caused by differing duplicate-record handling between databases and flaws in a custom metrics collection plugin. Across 10.3 million production query comparisons, ClickHouse reduced average query latency from 906 milliseconds to 33 milliseconds and p95 latency from 1,274 milliseconds to 116 milliseconds, while lowering database costs by roughly 85 percent and enabling schema changes, metadata joins, and broader future data-processing capabilities.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.