ClickHouse Releases PostgresBench: An Open Benchmark for Postgres Services

ClickHouse published PostgresBench on 20 June 2026, an open-source benchmarking project that evaluates more than 40 databases and makes its methodology fully transparent — queries, datasets, and raw results are all published publicly.

The project extends ClickHouse's existing benchmarking work. ClickBench, the company's OLAP benchmark — a tool for testing analytical database performance — already covers 40-plus databases using reproducible methodology. PostgresBench applies that same approach to the Postgres ecosystem, focusing specifically on the crowded market of managed Postgres services and Postgres-compatible alternatives, where vendors often claim performance advantages but independent comparisons are rare.

The key innovation is reproducibility. Because all queries, datasets, and results are public, any operator can rerun the benchmark against their own infrastructure and verify the published numbers. This matters in a benchmarking landscape where vendor-paid tests and hand-picked workloads have made many practitioners skeptical. The GitHub repository makes the entire methodology auditable — you can inspect not just the summary tables, but the actual code and logic underneath.

ClickHouse has built a track record of publishing benchmarks across different competitive areas. In February 2026, the company published results showing that pg_clickhouse — a Postgres extension that routes analytics queries to ClickHouse — delivered the fastest analytics performance among tested Postgres extensions. In January 2025, ClickHouse ran a JSON workload benchmark comparing ClickHouse against MongoDB, Elasticsearch, DuckDB, and PostgreSQL using a billion-document dataset. PostgresBench continues that pattern: each benchmark targets a specific competitive landscape and publishes methodology and results openly.

In May 2026, ClickHouse also released CostBench, which benchmarks cost-to-performance ratios across cloud data warehouses. Together, the benchmarking portfolio now spans OLAP performance (ClickBench), semi-structured data workloads (the billion-docs suite), cloud warehouse economics (CostBench), and now the Postgres service layer (PostgresBench). Each benchmark serves as both a reference tool for infrastructure decisions and a positioning statement for ClickHouse itself.

There is a practical caveat to keep in mind. ClickHouse has a direct commercial stake in some of these benchmarks — pg_clickhouse, for example, is ClickHouse's own product, and a benchmark showing it as the fastest also functions as marketing. That does not make the methodology flawed, but engineers considering PostgresBench results for infrastructure decisions should replicate the tests on their own representative workloads rather than treating published numbers as final. The fact that the methodology is fully reproducible and open source at least makes that verification straightforward and practical.

The Postgres service market has grown steadily as cloud providers — AWS, Google Cloud, Azure, Neon, Supabase, PlanetScale, and others — have expanded their managed database offerings. Many maintain Postgres protocol compatibility while using different storage engines, replication approaches, and query execution strategies. A benchmark covering 40-plus services with consistent methodology fills a real gap that has existed in this market. How widely PostgresBench is adopted and how actively ClickHouse updates it as the service landscape shifts will determine whether it becomes a standard reference for database teams.

The repository is publicly available on GitHub at ClickHouse/PostgresBench.