ClickHouse Managed Postgres

Unlock up to 10× faster Postgres performance with NVMe storage built for I/O-intensive workloads. Microsecond-level latency and millions of IOPS eliminate disk bottlenecks for high-throughput transactional applications.

Unlike traditional SSDs that cap out at a few hundred thousand IOPS, NVMe scales effortlessly, delivering lower tail latencies and faster system operations like VACUUM, checkpoints, and logical replication.

High availability is built in by design, with up to two dedicated standby instances reserved exclusively for HA. Read replicas are offered separately to preserve reliability. Automatic backups and point-in-time recovery (PITR) with WAL archiving ensure robust disaster recovery.

WalShadow (private preview) replicates Postgres to ClickHouse straight from the physical WAL, the same stream Postgres uses for its own standbys. There are no logical replication slots and no decoding load on your primary. In our benchmarks, commits showed up in ClickHouse in around 200 ms while sustaining 289K rows/sec. Schema changes (ADD, RENAME and DROP COLUMN), initial load, restart recovery and planned switchovers are all handled.

ClickPipes Postgres CDC remains available for initial migrations and as a fallback path. Replication from Managed Postgres to ClickHouse is free with either.

Query ClickHouse directly from Postgres using the pg_clickhouse extension. No proxies, no new query layer to manage. Postgres becomes a single interface for both transactions and analytics.

Queries are transparently pushed down to ClickHouse for maximum performance, supporting joins, filters, aggregations, and functions. Today, 14 of 22 TPC-H queries are fully pushed down, delivering 60×+ speedups.

Pushdown support continues to expand to CTEs, window functions, and more, enabling fast analytics from the Postgres layer while ClickHouse handles execution behind the scenes.