August 7, 2026
Fast database, faster comment war
Making Postgres 300x faster for analytics: batching, operator fusion, and SIMD
Postgres gets a turbo makeover, but the comments instantly started a licensing war
TLDR: pgrust says it made Postgres-style analytics up to 300 times faster, a huge claim that got people excited fast. But the comment section immediately split between hype over overdue improvements and a familiar open-source fight over whether the project’s license will scare users away.
A tiny database upstart just walked into the room claiming it made Postgres, one of the internet’s oldest and most beloved data engines, up to 300 times faster for big number-crunching jobs — and yes, the crowd absolutely lost it. The project, called pgrust, says its new version is not just faster than regular Postgres, but even edges past ClickHouse, a database famous for analytics speed. The pitch is simple enough for non-database people: old software was built for an era when hard drives were slow, but today the real battle is getting the computer’s brain and memory to stop wasting time.
But the real fireworks were in the replies. One of the hottest reactions wasn’t even about speed — it was about the license. A commenter bluntly called the project’s AGPL license “odd” and basically suggested someone could fork it into a more permissive version to win more fans. That is classic open-source drama: you announce a breakthrough, and the internet immediately argues about paperwork. Meanwhile, the author jumped into the thread like a nervous magician saying, “yes, it’s fast, but please also trust it,” pointing to formal checks and fuzz testing to calm fears that this is just a benchmark flex with sharp edges.
Elsewhere, the comments turned into a wishlist and therapy session for long-suffering Postgres users. People begged for comparisons to rival tools, demanded deep dives into the scheduler, and cheered the possibility of adaptive planning — basically a smarter way for the database to change its mind when reality doesn’t match the plan. The vibe was equal parts hype, skepticism, and “finally, someone said it.”
Key Points
- •The article says pgrust 0.2 is 10x faster than the previous pgrust release, 30% faster than Postgres on OLTP benchmarks, and 300x faster than Postgres on Clickbench.
- •It attributes a significant portion of the analytical speedup to a redesigned query engine, which the article says contributed roughly 10x of the 300x improvement.
- •The article argues that Postgres reflects an earlier era dominated by disk I/O, while modern workloads are more constrained by CPU throughput and memory bandwidth.
- •In a simple sum over 500 million floating-point values, the article reports about 20 seconds for Postgres versus 358ms for an equivalent Rust loop, while noting the comparison is not fully apples-to-apples.
- •The article explains that Postgres converts SQL into query plans and executes them with a Volcano model executor, then uses a miniature implementation to motivate pgrust's optimizations.