How pgrust made Postgres 300x faster for analytics

Postgres was built for 1980s disks, but updating its query engine makes analytical queries run 300x faster.

The team behind pgrust released version 0.2 with massive performance improvements. On Clickbench analytics benchmarks, pgrust ran 300x faster than standard Postgres and edged past Clickhouse. Overhauling the query engine accounted for a 10x speedup on its own.

Why it matters: Postgres was designed when reading from disk was the main bottleneck. Today, datasets live in RAM or fast NVMe drives, turning CPU overhead into the real bottleneck.

Know this: Standard Postgres uses the classic "Volcano model", which fetches data one row per function call. In a simple test summing 500 million numbers, Postgres took 20 seconds. A raw Rust loop did the exact same math in 358 milliseconds. Replacing row-at-a-time execution with modern CPU optimizations closes that gap.

Turns out 40-year-old database design choices leave plenty of performance on the table.