It was increasingly difficult for China Unicom to identify faults in its cloud core network, while the hard disks in its data center were below par, and links in its optical transport network were deteriorating. Meanwhile, customer expectations are ever higher, and they are intolerant of poor network coverage or a mismatch between devices, networks, and services. Fast troubleshooting to optimize user experience was essential, along with a comprehensive, accurate analysis of the ultra-large network and user data. Massive amounts of data are collected in 31 provinces and then aggregated centrally, facilitated by TM Forum’s Open APIs. The operator adopted Flink for low-latency stream processing. It uses Spark to analyze and process warehouse data and simplify the data model, deploying a data architecture that can converge batch and streaming data. The system mines petabytes (PB) of data to generate a unique data model for each subscriber, with Open APIs making data from many sources accessible.


Better efficiency


as cross-layer/cross-domain troubleshooting covers all faults on entire network

Service availability up by


enhancing user satisfaction

Better quality home Wi-Fi by


On-time fulfillment


Zhang Taoye

General Manager of Intelligence Network & Innovation Center

“China Unicom has built an intensive, platform-based, digitalized, intelligent, and agile Network Intelligent Operation System which is based on "Platform + Application" and under the Autonomous Network Architecture.”

Machine learning, human knowledge

China Unicom combines machine learning algorithms and expert knowledge to predict, demarcate, and locate cross-layer and cross-domain risks on the CT network and IT cloud for proactive maintenance. It converges monitoring, visualizes the topology, and locates faults based on cross-layer and cross-domain alarms. It makes network faults easier to manage by visualizing them.

Using AI, the operator built a scheduling system for its autonomous network to provide visible, manageable, and controllable dedicated services. They are delivered based on cross-domain intelligent networking, an intent engine, and resource data.

The command center enables universal self-optimization of network parameters, software resource allocation, and fault rectification. Users can provision and diagnose services, and match optimal packages with one click, in a day.

Data analysis, and the Forum’s return on investment calculator, helps China Unicom how understand networks and services contribute revenue in high-value areas, as well as for high-value users and potentially high-value customers.

TM Forum assets used

Find out more about how TM Forum membership can help your business

By signing up for this case study, you agree to the Privacy Policy