Architecture and Application Practice of JD Urban Spatio-Temporal Data Engine (JUST)

The talk introduces JUST, a spatio‑temporal data engine developed by JD Intelligent City Research Institute, explaining why traditional GIS and relational databases struggle with massive, high‑frequency, multi‑dimensional urban data.

Key challenges of spatio‑temporal data are its huge volume, complex structures, unique query patterns, and rapid updates. Existing solutions (PostGIS, Hadoop, Spark, GeoMesa, etc.) face scalability, efficiency, and usability issues.

JUST addresses these by providing a unified platform (JUST‑DB, JUST‑DM, JUST‑TS, JUST‑GIS, JUST‑Service) that integrates storage, indexing, mining algorithms, visualization, and service management, offering strong extensibility, high efficiency, good usability, and fast rendering.

The engine models spatio‑temporal data into six types (point vs. network, static vs. dynamic vs. temporal) and designs optimal indexes and algorithms for each, reducing storage by 85% and improving query speed by over 7× compared with vertical storage.

A novel horizontal trajectory storage scheme groups GPS points into sub‑trajectories, enabling compression and full‑trajectory queries, while a refined signature representation improves spatial precision.

JUST uses a NoSQL key‑value store (HBase) and distributed frameworks (Spark, Hadoop) with custom one‑dimensional encodings that separate time buckets from spatial codes, achieving sub‑5‑second range queries.

Two pre‑created SparkContexts managed by SparkJobServer reduce job‑launch latency and improve high availability.

Extensive experiments on real trajectory datasets show superior storage efficiency, query performance, and robustness, with results accepted at ICDE 2020.

Beyond the core engine, JUST provides ready‑to‑use spatio‑temporal mining algorithms, a SQL interface with optimizer, JDBC driver, and configurable API services that simplify backend development.

Visualization components (JUST‑GIS, JUST‑Studio) enable interactive map rendering, handling millions of points without lag, and support dynamic analysis such as real‑time traffic monitoring.

Real‑world deployments include COVID‑19 contact‑tracing in Beijing, hazardous‑material monitoring in Nantong, and fine‑grained road‑network reconstruction from courier GPS data, demonstrating the platform’s impact on public health, safety, and logistics.

JUST’s research contributions have earned multiple ACM SIGSPATIAL awards, papers in top conferences (ICDE, AAAI, KDD), over 30 patents, and certifications from Chinese authorities.