// the find
StarRocks/starrocks
The world's fastest open query engine for sub-second analytics both on and off the data lakehouse. With the flexibility to support nearly any scenario, StarRocks provides best-in-class performance for multi-dimensional analytics, real-time analytics, and ad-hoc queries. A Linux Foundation project.
StarRocks is an MPP OLAP database with a vectorized C++ execution engine and a Java frontend. It's aimed at teams running sub-second analytics on billions of rows, either directly in StarRocks tables or by querying Iceberg/Hudi/Delta Lake in place. Production deployments at Airbnb, Pinterest, and Coinbase give it more credibility than most self-published benchmarks.
The BE is C++ with SIMD-level vectorization (AVX2, NEON) — that's where the performance actually comes from, not marketing. The CBO is genuinely good enough to handle star schema joins without denormalization, which is the main reason teams pick it over ClickHouse. Shared-data mode (v3.0+) decouples compute from storage via S3-compatible backends, which changes the operational model significantly for cloud deployments. Intelligent materialized views auto-refresh on ingest and are transparently selected at query time — this actually works, unlike most 'automatic' MV implementations.
The FE (Java) and BE (C++) split means two completely different build toolchains, two different log systems, and two separate tuning knobs — debugging a slow query can take you across both. Building from source requires Docker and a significant amount of patience; the compile times for the C++ BE are brutal and not well-documented for first-timers. The community skews heavily toward Chinese-speaking users, so GitHub issues and StackOverflow coverage in English are thin — you will hit walls where the only answers are in Chinese forums. The 'world's fastest' claim comes from self-run TPC-DS benchmarks against configurations they tuned; real-world performance heavily depends on table model choice (duplicate vs aggregate vs primary key) and getting that wrong early is expensive to fix.