Guangdong Mobile's DevOps Turnaround: A Practical Operations Blueprint

1. Challenges

In 2018 GOPS Shenzhen, the financial and telecom sectors faced massive disruption from internet services such as WeChat, QQ, Alipay, and mobile payments, exposing a huge IT gap between traditional operators and internet companies.

Traditional “turn‑key” projects involve vendor‑delivered packages deployed on‑site, resulting in high operation costs, low efficiency, and poor quality. In contrast, internet firms adopt a self‑develop‑self‑operate DevOps model that delivers quickly and reliably.

2. Choosing a DevOps Path

Enterprises can either rely on outsourcing (“blood transfusion”) or pursue self‑development (“blood replacement”). Guangdong Mobile opted to cultivate internal “genes” and continuously improve its own “blood‑making” capabilities.

3. Technical Practice

3.1 Selecting a Comprehensive Framework

Two reference models were recommended: the DevOps “Dao‑Fa‑Shu‑Qi” community framework and the China Academy of Information and Communications Technology’s “R&D‑Operations Integrated Capability Maturity Model”.

“Any improvement made outside the bottleneck is an illusion.” – Eliy Goldratt, *The Goal*

3.2 Focusing on Six Core Processes

From the 42‑process model, Guangdong Mobile prioritized six processes to address immediate pain points:

Story and task scheduling – managing stakeholder expectations.

Deployment and distribution model – improving version‑release efficiency.

Application monitoring – accelerating fault detection and root‑cause analysis.

Incident response and handling – streamlining failure remediation.

High‑availability planning – enhancing system reliability.

Metrics – quantitatively evaluating DevOps transformation.

3.3 Detailed Process Implementation

Story and task scheduling : Establish annual resource plans, allocate “train tickets” for each demand, and reserve 10 % buffer capacity to balance workload across quarters.

Deployment and release model : Introduce a pre‑production environment and a two‑stage deployment pipeline to avoid cross‑system night‑time releases.

Application monitoring : Deploy standardized data collection, use Oracle RUEI for performance management, and build end‑to‑end business monitoring that feeds into unified log analysis and AIOps.

Incident response : Script routine fault‑handling tasks, gradually introduce intelligent automation, and capture common failure patterns from vendor interviews.

High‑availability planning : Apply cloud‑native concepts such as multi‑availability‑zone deployment, anti‑affinity rules, and layered HA (network, host, SaaS, storage, application).

Metrics : Define a DevOps evaluation system covering CMDB, HA, alarm & log, repair, and continuous delivery, and codify database programming standards.

4. Management Insights

Transformation requires “seizing the trend”, “leveraging the trend”, and “creating the trend”. Strategies include aligning with NFV initiatives, inviting external experts, organizing internal DevOps competitions, and fostering a culture where IT staff can write code.

5. Conclusion

DevOps cannot be copied wholesale; each organization must adapt the model to its leadership style, talent, and vendor capabilities. Guangdong Mobile’s goal is to reduce release cycles from months to days, improve operational efficiency, and relieve night‑shift burdens for staff.

6. Q&A Highlights

Key questions addressed included the business value of DevOps, the balance between self‑development and outsourced components, talent cultivation, and practical steps for sustaining the transformation.