JD's J-one Unified Release Platform: Architecture, Features, and Recent Enhancements

Since July 2016, JD's unified release platform J‑one has supported more than 90% of the company's backend services, handling everything from compilation and configuration to deployment, and offering a range of auxiliary tools. Recently, in partnership with public‑cloud provider Jingcheng, J‑one added a new image‑deployment feature that greatly improves release efficiency.

To keep up with JD's rapid business expansion, the original single‑machine deployment architecture was redesigned into a horizontally scalable system that can automatically expand or shrink based on load, eliminating queuing bottlenecks for developers.

The main goals of the platform’s upgrades are to increase developer deployment efficiency, raise deployment success rates, and enrich the platform with troubleshooting utilities.

Traditional deployment at JD involved lengthy manual steps such as VM provisioning and environment preparation, which slowed down releases, especially under tight resource constraints. The new J‑one platform integrates public‑cloud resource requests and leverages containers for rapid scaling, maximizing resource utilization.

J‑one now supports image deployment: a deployable image is built once, tested offline for consistency with production, and then released, synchronizing logs and monitoring automatically. This reduces the typical three‑hour traditional deployment to a matter of minutes.

The platform also implements a full CI/CD pipeline, handling diverse application languages and middleware. It offers customizable container compilation using Kubernetes auto‑scaling, providing on‑demand compile images for C++, Go, Node.js, etc., cutting compile time by at least half with only ten machines handling tens of thousands of daily builds.

Automation tools added in the latest upgrade include package dependency comparison, dependency management, application‑level dependency analysis, system/application export, configuration restart control, and self‑service Nginx file operations, all aimed at speeding up business response and reducing manual intervention.

Performance improvements have increased deployment speed by five times, enabling parallel releases to 300 machines within about five minutes. Ongoing efforts focus on strengthening infrastructure, stability, and reliability to deliver a robust, user‑friendly platform.