Scaling DevOps at Huawei: Secrets Behind a Massive Cloud Platform Transformation

Background

2017 marked the first year DevOpsDays entered China, with conferences in Beijing, Shanghai, and Shenzhen. The speaker, a Huawei product manager, shares experience of large‑scale DevOps transformation.

Huawei’s R&D Evolution

Huawei progressed through three software‑engineering eras: a workshop era before 1998, a heavyweight process‑control era after adopting IBM’s IPD in 1999, and the agile/lean/DevOps era since 2008, now called the “special forces” era.

Definition of DevOps at Huawei

Huawei defines DevOps not only for software development but as a catalyst for business‑model change, enabling traditional manufacturing firms to shift from product to service models and supporting cloud and ICT transformation.

DevCloud – A One‑Stop DevOps Platform

DevCloud is Huawei’s cloud‑native DevOps platform, built as an internal startup. It grew from a two‑person team to over a hundred engineers, following typical startup stages: market discovery, MVP, product‑market fit, and scaling.

Full‑Function Teams and Organizational Change

Huawei adopts full‑function (cross‑functional) teams that own feature planning, design, development, testing, delivery and operation. Teams are typically 10‑15 people per service, with roles such as product manager, developers, testers, architects, UED designers and operations staff.

Management Practices

Iterations are one week long, with two acceptance checks per iteration. The team uses a value‑stream board for each service, enabling transparent tracking of design, development, testing and release.

Scaling Agile with SAFe

Most product lines use the SAFe framework (product‑level agile). However, DevCloud grew without it, showing that large‑scale frameworks are optional.

Cultural Foundations

Huawei’s “long‑term self‑criticism” culture drives quality retrospectives and post‑iteration reviews, fostering continuous learning and improvement.

Transformation Strategy

The roadmap follows John Kotter’s eight‑step change model adapted for DevOps: create urgency, secure top‑level endorsement, set quantitative goals, pilot, roll out, spread horizontally, and embed the new practices into culture.

Conclusion

The speaker emphasizes that DevOps transformation requires aligning architecture, organization, and culture, and that incremental, data‑driven improvements enable large‑scale success.