How ICBC Built a High‑Performance DevOps Platform to Accelerate Financial Innovation

DevOps Platform Evolution

Since 2009 the bank has progressively standardized build and deployment tools, introduced continuous integration with Jenkins in 2013, and launched a dedicated DevOps platform in early 2018. The platform integrates development, testing, and operations to automate version construction, quality assurance, delivery, hand‑over, and production deployment.

Toolchain Overview

Developer Portal : A unified web entry that aggregates all toolchain functions for developers.

Continuous Delivery Pipelines : Jenkins‑based pipelines that chain code commit, static analysis, build, deployment, and automated testing.

Operations Platform : Centralized console for production operations, application monitoring, and capacity planning.

Automated Pipelines

The end‑to‑end pipeline reduces a single‑application cycle to roughly 2 hours , with the actual deployment step limited to 0.5 hour . The pipeline consists of the following stages:

Build Submission Pipeline : Triggered on every code commit. Executes static code scans and unit‑test coverage checks. If any quality gate fails, the merge is blocked.

Continuous Integration Pipeline : Runs after a successful merge. Performs another round of scans, builds the artifact, deploys to a test environment, and executes automated functional tests. A smoke‑test success criterion must be met before the version can proceed.

Hand‑over & Deployment Pipeline : Automates the final hand‑over to operations and the production deployment, including environment‑specific configuration.

Quality Gates : Enforce “first‑time‑right” standards for code quality, test coverage, and smoke‑test pass rate. All gates must be satisfied before promotion.

Operations Platform : Provides a unified dashboard for monitoring deployed services, managing incidents, and planning capacity.

Standardization Measures

After two years of DevOps practice the bank identified gaps in code management, version building, and production deployment. In 2020 a set of technical specifications was introduced and the toolchain was refactored to improve standardization and automation.

Self‑service Efficiency : Teams own the full lifecycle (“who develops, who operates”), reducing operational bottlenecks.

Standardized Release Units : Decoupled business and application architectures; unified repository layout, build scripts, deployment directory structures, and environment definitions.

Infrastructure‑as‑Code for Build/Deploy : Build and deployment processes are described with layered yaml files, creating separate “build‑cloud” and “deploy‑cloud” modules.

Component‑Based Pipelines : Defined component integration standards enable horizontal customization of workflow sequences and vertical customization of individual tasks.

Metrics Platform : Aggregates data from all subsystems to provide role‑based, multi‑dimensional visual dashboards that drive continuous improvement.

Full‑Feature Views : End‑to‑end visualizations of the delivery pipeline for all stakeholders.

Future Direction

The current toolchain still follows existing workflows, but as the platform matures manual control points will be eliminated. The long‑term goal is to empower product teams to deliver usable services directly, deepen Dev‑Ops collaboration, and adopt Site Reliability Engineering (SRE) principles. This will create an automated, self‑service, intelligent DevOps platform that lowers software delivery barriers and improves overall IT efficiency.