From Waterfall to AIOps: How One Ops Leader Transformed Zhejiang Mobile’s IT

From Waterfall to AIOps: Opportunities and Challenges

Zhejiang Mobile’s IT is divided into B, M, and O domains, where B supports business systems, M handles management information, and O runs network operations. Fifteen years ago, Fang Wei started in B‑domain customer‑service system maintenance, later moving to development, architecture governance, and demand management.

At that time the team followed a waterfall model: requirements were handed to partners, developed, tested, and finally deployed, often causing repeated releases and information loss. The process was passive, with teams shifting blame rather than owning end‑to‑end outcomes.

Adopting Agile and DevOps

Around 2013, driven by a customer‑experience focus, Fang Wei introduced agile practices, starting with the mobile “self‑service” portal. The team defined a four‑step agile framework: loosen the soil (mindset change), try, solidify, and promote, eventually forming the Fast+ agile standard and earning a national innovation award.

Subsequently, the team embraced DevOps, merging development, testing, and operations into a single department. Recognizing rapid cloud evolution, they built a container‑centric DCOS platform, integrating cloud services with DevOps, simplifying B‑domain development and testing through micro‑services and containerization.

Building the Cloud Management Platform

In 2017, Fang Wei led the creation of the AD Cloud (Agile Delivery Cloud), later renamed Kunlun, forming Zhejiang Mobile’s “Qiantang Cloud.” The platform’s core is a CMDB that unifies IaaS and PaaS resources, enabling business‑centric cloud management where resources can be provisioned within ten minutes.

Initial CMDB challenges included data decay due to manual processes. The team automated data collection, eliminated manual interfaces, and later used a graph database to capture relationships between network elements, addressing the most critical aspect of cloud‑based operations.

Multi‑Level Fault Isolation (定界)

The team defined a four‑level fault isolation hierarchy to accelerate incident handling:

Level 1: Is the issue in the application or the cloud platform?

Level 2: Which cloud component is problematic?

Level 3: Which functional module of the component fails?

Level 4: On which node does the failure occur?

This systematic approach paved the way for AIOps initiatives.

AIOps: Transforming Operations

Fang Wei views AIOps as both an opportunity and a challenge. While full automation may reduce traditional ops roles, it elevates operations to a strategic, AI‑driven discipline requiring data engineering, algorithm expertise, and architectural insight.

He emphasizes that Zhejiang Mobile’s strong AI and big‑data foundation, combined with a culture of experimentation, enabled early AIOps adoption.

Differences Between CT and IT Domains

In CT (communication technology), systems are vertical silos without cloud or agile practices, demanding persistent connections (e.g., voice calls). In contrast, IT systems focus on interaction, data persistence, and analytics, allowing layered, decoupled architectures.

CT cloudification faces three decoupling challenges: soft‑hard (software/hardware), platform‑application, and application‑cloud‑hardware layers. Achieving cloud‑native CT can enable dynamic, high‑quality network services such as 8K video streams and 5G edge computing for low‑latency, high‑bandwidth scenarios.

Impact of DevOps Standards

As a core author of the China Information and Communication Research Institute’s DevOps standard, Fang Wei stresses that DevOps bridges development and operations, prioritizing customer experience and business value.

Standardized DevOps practices help partners quickly adopt agile workflows, fostering a business‑centric mindset across the organization.

Advice for Young Operations Professionals

Fang Wei recommends continuous learning to become an expert, and staying abreast of emerging technologies such as AI and big data. He also shares a six‑step communication framework: state the current situation, identify problems, pinpoint root causes, set change goals, propose solutions, and define expected outcomes.

Overall, his 2‑hour interview provides valuable insights into the evolution from traditional waterfall ops to modern, AI‑enhanced, cloud‑native practices.