GeaBase: China’s First Financial‑Grade Distributed Graph Database – Architecture, Performance, and Real‑World Applications

GeaBase is the first Chinese financial‑grade distributed graph database, developed by Ant Financial to address the limitations of relational databases for deep‑relationship queries such as multi‑degree social‑network recommendations and fraud detection.

The system stores data as nodes and edges, enabling fast traversal of complex relationships. Benchmarks from the book Graph Databases show that while relational databases struggle beyond depth 5, GeaBase completes the same queries in 2.132 seconds, far outperforming MySQL and Neo4j at lower depths.

GeaBase’s architecture is built for massive scale: it supports billions of vertices and hundreds of billions of edges, offers multi‑cluster deployment, automatic failover, and a self‑developed query language that is extensible, supports custom functions, integrates dynamic languages like Python and C++, and remains compatible with Gremlin.

Key application scenarios include real‑time risk control, anti‑money‑laundering, fraud detection, community detection, and recommendation engines across Ant Financial’s products (Alipay, Ant Wealth, etc.) and Alibaba Group services (Taobao, Tmall, etc.). The platform can process online queries in milliseconds and offline graph computations for model training.

Since its first release in early 2016, GeaBase has been stress‑tested during major events such as the 2017 Spring Festival Red‑Packet campaign, proving its ability to handle peak loads of millions of transactions per second. The product has now entered a free public‑beta phase and is being prepared for commercial cloud‑based offerings.