How StarRocks Redefines Lakehouse Architecture with Ultra-Fast Unified Analytics

StarRocks Background

StarRocks is positioned as an ultra‑fast and unified OLAP engine, aiming to dramatically improve query efficiency in analytical scenarios.

Positioning: Speed and Unification

Speed : From version 1.0, StarRocks focuses on extreme performance using CBO and vectorized execution.

Unification : Since version 2.0, it unifies core OLAP scenarios—multidimensional analysis, real‑time analytics, high‑concurrency queries, and ad‑hoc queries—under a single technology stack, reducing operational overhead.

Community and Evolution

The StarRocks community is highly active, with close collaboration with Alibaba Cloud for nearly three years, contributing to rapid feature iteration.

StarRocks 3.x Features

Storage‑Compute Separation

Version 3.x separates storage from compute: data is stored in external systems such as OSS or HDFS, while compute nodes (CN) become stateless, improving flexibility, scalability, and cache utilization.

Benefits

Storage cost reduction of 70‑80% by using single‑replica OSS storage.

Elastic compute scaling via Warehouse management, enabling on‑demand resource allocation.

Improved reliability through OSS high‑availability architecture.

Resource Isolation

CN nodes can be grouped into independent resource units (Warehouses), preventing resource contention among different workloads.

Multi‑Warehouse

Provides physical isolation of CPU, memory, network, and I/O for different workloads, with elastic scaling (planned release June 2024).

Lakehouse Analysis

StarRocks supports reading and writing to data lake formats (Hive, Iceberg, Paimon) via a unified catalog, enabling seamless lake‑warehouse fusion and achieving 3‑5× performance over Trino/Presto.

Trino Compatibility

By setting set sql_dialect = "trino", StarRocks can parse Trino SQL with ~90% compatibility, allowing smooth migration from Trino/Presto.

Materialized View Acceleration

Materialized views provide transparent query acceleration for both lake and traditional OLAP scenarios, reducing the need for complex ETL pipelines.

EMR Serverless StarRocks Product

The fully managed EMR Serverless StarRocks offers:

Optimized performance for primary‑key tables and point queries (2‑3× faster than Doris).

Unified lake support, including integration with Paimon and other lake formats.

Enhanced security and RBAC, simplifying permission management for both internal and external tables.

Seamless integration with DataWorks for real‑time data ingestion and batch loading.

Zero‑maintenance SLA with automatic upgrades, health reports, and visual SQL editor.

Instance Management and Tools

Management console provides instance scaling, storage expansion, network configuration, health diagnostics (slow SQL, hot tables), visual import tasks, metadata visualization, and a built‑in SQL editor for ad‑hoc queries and development.