48.8kThe first open benchmark for data warehouse cost-performance
Most benchmarks measure query speed. CostBench measures performance per dollar across the full path: ingest, preparation, and reads.
A 50ms query can still be slow and expensive to serve
Before data can be queried, it must be ingested, organized, and kept query-ready. Sorting, indexing, materialized view maintenance, and freshness all take time and resources.
Traditional benchmarks usually measure only the read step.
CostBench measures the whole path.
Full-path real-time analytics results
CostBench measures the complete real-time analytics path: fresh data in, query-ready data maintained, fast answers out.
In the first full-path benchmark, we used a financial-services workload based on NBBO stock-market quotes: 1M quotes/sec, 100B rows in ~28 hours, with dashboard and drill-down queries running continuously.
The result: once the whole path is measured, systems separate.
ClickHouse leads on cost-performance across the full path
The score combines the cost of keeping fresh data query-ready with the runtime needed to serve continuous queries. Lower is better.
More results coming soon
Lower is better. Relative to ClickHouse at 100B rows.
Where the savings come from
- Ingest costs up to 82% less
than Snowflake and Databricks. - Preparation costs up to 65% less
than Snowflake and Databricks. - Read costs up to 83% less
than Snowflake and Databricks.
Curious about read-side only?
We benchmarked this across key competitors and delivered cost-performance at scale.
Lower is better. Relative to ClickHouse (1x).
Open methodology
Workloads, pricing assumptions, configurations, and benchmark code are all available publicly, making it possible to inspect, validate, and reproduce every result.
Open source
Benchmark code and tooling are publicly available.
Real pricing
Uses publicly available vendor pricing models.
Reproducible
Same workloads, same datasets, same methodology.
Transparent
Configurations, assumptions, and results are documented.
Run it yourself and validate our numbers.
Beyond the benchmark
All proven by benchmarks that can be reproduced by anyone.
At ClickHouse, we measure performance relentlessly - believing every millisecond matters. We continually push to make queries faster.
We back up our claims with public, reproducible benchmarks like ClickBench and JSONBench. You can explore our results directly and compare ClickHouse against other technologies before deciding how to power your analytics.