AliSQL: Alibaba's open-source MySQL with vector and DuckDB engines

Alibaba drops a souped‑up MySQL with DuckDB and AI search vibes — Postgres fans sound off

TLDR: Alibaba open-sourced AliSQL, a MySQL spin with built-in DuckDB analytics and planned AI-style vector search. Commenters are split: some hail a big productivity win, while Postgres fans ask why not use extensions like pg_duckdb—sparking a lively MySQL vs Postgres showdown.

Alibaba just open‑sourced AliSQL, a MySQL remix that bundles in DuckDB for quick, spreadsheet‑style analytics and teases future AI vector search (think: math fingerprints for text/images) with up to 16,383 dimensions using a fast “find‑the‑nearest‑thing” algorithm. Translation: do everyday app stuff and data crunching in one place. The crowd? Loud. The repo landed and HTAP—hybrid transactional/analytical processing, mixing live app data with analytics—had people cheering. One fan called the embedded analytics a productivity “massive win”, while another asked how it stacks up against pg_duckdb, the Postgres add‑on doing similar tricks. The Postgres crew flexed: “extensions > forks,” pointing to clean integration via Postgres’ plugin system. Meanwhile, AliSQL stans fired back: native engine, familiar MySQL, fewer moving parts. Keyframe even dropped a spicy crossover: “MySQL + DuckDB + vectors = Vespa‑lite?” The drama escalated when a commenter scolded the laugh‑react crowd—no giggles allowed—as the MySQL vs Postgres turf war warmed up. Behind the memes (“Duck season? DuckDB season!”) were real questions: will planned vector search and faster schema changes beat the Postgres extension ecosystem, or is this just a shiny fork? Curious minds dove into AliSQL’s DuckDB page (link) to see if the hype matches reality.

Key Points

  • AliSQL 8.0.44 (LTS) is an open-source MySQL 8.0.44-based fork used in Alibaba Group’s production.
  • AliSQL integrates DuckDB as a native storage engine to enable lightweight analytics with MySQL-like operation.
  • Planned vector storage supports up to 16,383 dimensions using an optimized HNSW algorithm for ANN search via SQL.
  • Roadmap includes DDL, crash recovery (RTO), and replication optimizations (e.g., Binlog Parallel Flush, Binlog in Redo).
  • Build prerequisites and instructions are provided, with GPL-2.0 licensing and support via GitHub and Alibaba Cloud RDS.

Hottest takes

"having an embedded column database for analytics in your traditional db is a massive win for productivity + operations simplicity." — dzonga
"Curious how it stacks up to pg_duckdb." — linuxhansl
"HTAP is here!" — jimmyl02
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.