How JD Scaled Its Order Fulfillment Center to Handle Millions of Orders

Importance of OFC

During the 2014 618 promotion, JD faced unprecedented order volumes, making the Order Fulfillment Center (OFC) a critical link between user checkout and warehouse production. OFC converts user orders into production orders for downstream systems, ensuring timely delivery and maintaining customer experience.

Formation of OFC

After JD’s website launch in 2003, rapid e‑commerce growth led to a proliferation of business systems. By 2011 a small team was created to transfer order data between systems, eventually handling both customer and non‑customer orders, formalizing the OFC.

.Net to Java Migration

To align with JD’s technology strategy, the legacy .Net order‑splitting and transfer services were rewritten in Java, followed by phased traffic switching and full deployment by early 2012.

211 Order Fulfillment Improvement Project

The "211" project upgraded eleven systems, including order transaction, pipeline, splitting, transfer, task, OFC, pre‑sorting, label, tax, invoicing, and WMS. In 5066 man‑hours the team reduced order‑to‑warehouse time from over two hours to under five minutes, introducing Zookeeper, CXF timeout settings, Log4j multi‑Tomcat configurations, Oracle Exadata, and MySQL partitioning.

SOP Joint Order Project

In 2013 JD launched a joint order (SOP) feature allowing a single checkout for self‑operated and POP merchant items. The team built a high‑performance splitting service, introduced caching, asynchronous processing, and a deduplication mechanism to prevent duplicate order submissions.

Transfer Architecture Upgrade

The transfer system was re‑engineered with asynchronous processing, parallel data handling, caching for hot data, and flow‑control to protect downstream systems during traffic spikes. Operational tools and configurable switches were added to mitigate risk during deployments.

From 618 to Double‑11

Order volume grew exponentially from 2012 onward, demanding a shift from database‑centric control to a workflow‑centric Process Control Center, improving data consistency and enabling state‑machine driven orchestration.

Supporting Massive Order Processing

Core OFC services were built with horizontal scalability and distributed deployment, allowing rapid capacity expansion for peak events like 618.

Ensuring Data Consistency

A unified main‑process flow and state machine were introduced to coordinate across logistics and finance domains, reducing data divergence and enabling graceful degradation during high load.

Operational Support

Monitoring, alerting, and a unified analysis platform were established to detect bottlenecks, isolate failures, and automate recovery, while a SOA governance platform maps system dependencies and enforces SLA compliance.

Big Data Foundations

Key principles include handling massive data volumes, defining SLA metrics per subsystem, prioritizing critical flows, and implementing flexible degradation strategies.

System Protection

Targeted flow control and traffic shaping protect downstream systems from overload, complemented by unified capacity monitoring and fast‑reject mechanisms.

Distributed System Design

Tasks are sharded into groups that can be scaled independently; a distributed task queue (Redis) feeds a task engine that orchestrates service calls, with an exception queue for failed tasks. Zookeeper provides dynamic workflow configuration and automatic throttling/acceleration based on system health.