Core Viewpoint - JD Logistics is facing complex challenges in the field of intelligent monitoring for its large-scale warehousing system and is actively exploring solutions to meet the growing demands of warehouse management [1][10]. Group 1: Challenges Faced - JD Logistics operates approximately 600 large warehouses with a total area exceeding 15 million square meters, and the increasing international warehousing has led to unstable network environments [1][12]. - Rapid business development has resulted in frequent asset changes, making the opening and closing of warehouses a norm [1]. - The monitoring system must cover thousands of machines and applications, including traditional devices, Docker containers, and cloud hosts, which adds to the complexity [1]. - The inconsistency of deployment environments and the presence of multiple application release environments further complicate monitoring [1]. - New challenges include the scattered application scenarios of AIOPS, varying maturity levels, lack of depth and specialization in monitoring metrics, incomplete data sources, declining reliability of the Configuration Management Database (CMDB) due to frequent asset changes, and a shortage of operational experts and versatile talents [1]. Group 2: Solutions Implemented - JD Logistics is actively building a large-scale monitoring system solution, defining the purpose and value of monitoring to cover multiple aspects such as faults and performance [2]. - The company has planned a monitoring operation system that includes tools, intelligence, data, and platform layers, and has designed key components such as access services, API gateways, and monitoring platforms [2]. - Goals for the system include timeliness, accuracy, precision, and compatibility, with a capability maturity model introduced to guide system construction [2]. - To ensure the reliability of the CMDB, JD Logistics employs various methods such as automatic discovery, message synchronization, process automation, and manual maintenance [2]. Group 3: Intelligent Monitoring Practices - In the practice of intelligent monitoring aimed at AIOPS, JD Logistics utilizes various anomaly detection methods, including adjacent moment comparison, year-on-year and month-on-month analysis, baseline-based detection, and Holt-Winters forecasting [3]. - The implementation of a call chain function supports distributed transaction tracking, suitable for large-scale cluster monitoring needs [3]. - The event processing engine combines rule and execution engines to automate alert handling and allow for manual intervention [3]. - An intelligent knowledge base linked with the ticketing system accumulates operational knowledge, providing convenient search and intelligent customer service [3]. - Additional practices include fault snapshots, network detection models, trend forecasting, and visualization to enhance the system's intelligence [3]. Group 4: Technical Architecture - The intelligent monitoring system of JD Logistics encompasses multiple stages, including data collection, analysis, decision-making, and processing, achieving comprehensive handling and application of monitoring data [6]. - The company emphasizes system compatibility by integrating various platforms to ensure the broad applicability of the monitoring system [6]. - These efforts not only enhance the intelligence level of the system but also provide strong support for warehouse management [6]. Group 5: Future Directions - JD Logistics plans to continue optimizing its technical architecture, product architecture, and organizational structure, focusing on AI and algorithm technologies to build a more complete intelligent operation system [10]. - The company will further explore AIOPS application scenarios to enhance the system's predictive capabilities, intelligent alerts, and automated processing abilities, aiming to better identify, resolve, and mitigate issues for more efficient and intelligent warehouse management [10].
京东物流超大规模仓储：智能监控的挑战、实践与未来规划