JD Logistics Warehouse Database Architecture, Automation, Performance Optimization, and Scaling

JD Logistics provides a fast shopping experience backed by a robust warehouse management system (WMS) that relies on a sophisticated database architecture.

The WMS database can be deployed in two main modes: a local mode where the application and database reside in the same warehouse to reduce latency, and a centralized mode where a single WMS cluster in an IDC serves multiple regional warehouses, improving resource utilization but introducing network latency risks.

To manage such a large‑scale system, JD built the UDBA automated DBA platform, which provides daily MySQL slow‑log analysis, performance monitoring, routine inspections, SQL training for developers, and a knowledge base of common issues.

Performance optimization is achieved through real‑time monitoring, collaborative SQL tuning with developers, periodic inspections, and training on MySQL best practices. A notable case involved diagnosing frequent InnoDB deadlocks by examining SHOW ENGINE INNODB STATUS and working with developers to adjust transaction logic.

The fault‑self‑healing system automatically switches primary/replica nodes and restarts failed components, leveraging MHA high‑availability techniques and custom scripts that adjust innodb_flush_log_at_trx_commit and sync_binlog settings to recover replication quickly.

For data archiving, JD adopts a three‑month retention policy for production tables and a one‑year retention for reporting tables, moving older data to an IDC using CockroachDB, which offers high‑concurrency writes and dynamic scaling.

Finally, JD upgraded the warehouse MySQL instances from 5.5 to 5.7, gaining significant performance gains, multi‑threaded replication, GTID support, and online DDL, while also refreshing hardware with SSDs and higher‑spec CPUs to improve overall availability.