ClickHouse Expands PostgresBench with High Availability Performance Analysis
Key Takeaways
- ▸ClickHouse released comprehensive HA benchmark results addressing the most common community feedback on PostgresBench
- ▸Two competing architectures dominate: shared-nothing (isolated compute/storage) vs. shared-storage (separated layers) with different durability/performance profiles
- ▸Customers must balance recovery point objectives (RPO) with performance—zero data loss requires additional compute resources and overhead
Summary
ClickHouse has published new results from PostgresBench, its open-source benchmark tool, measuring the performance cost of high availability configurations across managed Postgres services including AWS RDS, AWS Aurora, Crunchy Bridge, Neon, and ClickHouse's own managed offering. The benchmark compares two fundamental architectures: shared-nothing services that rely on native PostgreSQL streaming replication, and shared-storage services that separate compute and storage layers. The results help practitioners understand the trade-offs between different HA configurations—from asynchronous replication with faster performance but lower durability guarantees (RPO > 0) to synchronous quorum replication supporting zero data loss (RPO = 0). ClickHouse's own offering provides flexible configuration options (0, 1, or 2 standbys) to match workload criticality, while the benchmark methodology aims for transparency and reproducibility across vendors.
- Open-source methodology enables reproducible vendor comparison and drives vendor accountability in managed database services
- ClickHouse provides flexible HA tiers from single-standby asynchronous to two-standby synchronous quorum configurations
Editorial Opinion
This benchmark fills a critical gap in the Postgres managed services landscape. By transparently measuring the performance costs of different HA architectures, ClickHouse gives practitioners concrete data to guide infrastructure decisions rather than relying on vendor marketing claims. The willingness to benchmark competitors fairly—and to expose the engineering trade-offs in your own product—builds trust and elevates the entire ecosystem's accountability.



