Striding Towards

the Intelligent World

White Paper

ADN
Building Self-fulfilling,
Self-healing,
and Self-optimizing
Autonomous Networks

Building a Fully Connected, Intelligent World

FOREWORD

Digital technologies, such as 5G, cloud, and AI, fulfilling intelligent networks with efficient O&M
are witnessing groundbreaking development capabilities.
beyond established boundaries. Human society
is evolving towards an intelligent world at an IP technology reconstructed the forwarding
unprecedented speed. Networks are becoming architecture of the communications network
increasingly significant as the ICT foundation for two decades ago. Cloud technology profoundly
achieving intelligent connectivity of everything. t ra n s f o r m e d t h e n e t w o r k m a n a g e m e n t
and control architecture a decade ago. The
The pandemic has made work from home+office, next decade will witness the application of
online+offline education, and virtual+real social AI technologies across all layers in network
entertainment the new norms. Home networks architectures. This will facilitate the evolution
not only fulfill daily life and entertainment towards high-level network autonomy and spark
purposes but also meet the requirements for a revolutionary upgrade for the communications
working from home, learning, and production. industry. AI will transform automated networks
These requirements significantly increase the to intelligent networks, machines assisting
demand, usage duration, and quality standard for humans to humans assisting machines, and on-
home networks. demand readiness to real-time awareness and
optimization of service experience.
Meanwhile, global enterprises are accelerating
digital transformation. This trend is transforming Achieving full network autonomy calls for
net works from mere support sy stems to tremendous breakthroughs in key technologies
important components in production systems. (such as hyper-converged awareness, large-scale
Networks must feature on-demand readiness, real-time simulation, and associated cognitive
provide deterministic SLA assurance, and deliver decision-making), as well as close collaboration
an optimal service experience. Enterprises are among industry players in terms of technical
shifting towards network as a service (NaaS) in architecture, interface standard, evaluation
terms of deployment, delivery, and management. system, and business model.

The focus of Huawei's unceasing innovations Only our joint endeavor can overcome all the
is directed at the autonomous driving network daunting challenges along this journey.
(ADN), ubiquitous cloud services, pervasive AI,
and low-carbon development benefiting from We firmly believe that a bright intelligent world is
digital technologies. As a crucial strategy of around the corner.
Huawei Communications Network 2030, ADN
is designed to deliver a new digital network
service experience featuring zero-wait,zero- Lu Hongju
touch, and zero-trouble to customers, and President of Huawei's General
enable self-configuration, self-healing, and self- Development Dept
CONTENTS
Chapter 2 06
ADN Trends and
Suggestions
2.1 Trend 1 2.6 Trend 6
2.2 Trend 2 2.7 Trend 7

Executive 01 2.3 Trend 3 2.8 Trend 8

2.4 Trend 4 2.9 Trend 9
Summary and 2.5 Trend 5 2.10 Trend 10
Keywords

Chapter 3 36
Huawei ADN Solution

3.1 Architecture and Features of

Huawei ADN Solution

3.2 Panorama of Huawei ADN

Chapter 1 02 Solution

Overview of the
AN Industry

Chapter 4 65
Outlook and
Initiatives
ADN

Executive Summary and Keywords

1 Executive Summary

Over the past decade, 4G and cloud have profoundly transformed the way we communicate
and live. Software-defined networking (SDN) and other network technologies continue to
facilitate application-driven network automation. Instant communication technologies allow us to
communicate anywhere anytime and have dramatically changed production methods. The next
decade will be a golden decade for a thriving intelligent era, with networks fully evolving towards
5.5G by 2025. The transformation from 4G to 5.5G calls for a self-fulfilling, self-healing, and self-
optimizing Autonomous Networks (AN) that make all things sensing, connected, and intelligent.
In this white paper, Huawei and industry partners provide insights into the trend of network
automation and intelligence, and offer suggestions on AN evolution to promote innovation and
development across the AN industry.

2 Keywords

digital transformation, AN, AI, ADN, digital twins, intent driven

1
01
Overview of the
AN Industry

2
 ADN

We are living in times of significant changes As the foundation of digital infrastructure,

and uncertainty, which have been the prelude network connectivity is playing an increasingly
to every single industrial revolution. The important role in promoting sustainable

development and large-scale application of 5G development and digital transformation in

various industries.
technologies have created a fully connected,
intelligent world. Booming digital technologies
The total number of global connections is
are revitalizing the economy, while the long-
likely to reach 200 billion by 2030, as networks
lasting COVID-19 pandemic further accelerates
will evolve from connecting tens of billions of
the digitalization of products and services. Over people to connecting hundreds of billions of
170 countries have released their national digital things. New services such as next-generation
strategies. Digital development has become human-machine interaction (AR/VR/XR),
an important consensus around the world. integration of housing and transportation,
According to McKinsey, the pace of digitalization industrial interconnection, satellite broadband

is seven years ahead of schedule on a world interconnection, and AI computing raise new
requirements for network connections. All this
scale, and 10 years ahead of schedule in the
sets the direction of network development,
Asia-Pacific region. The speed of digitalization
i.e., a stereoscopic cubic broadband and green
for operator and enterprise services is 20 to
network that features native intelligence, security
25 times faster than expected. Digitalization
& trustworthiness, deterministic experience, and
is no longer a difficult task as it was thought
integrated sensing and communication.
to be. Work from home+office, online+offline
education, and virtual+real social entertainment 5G has resulted in an increasing network scale and
will gradually become the new norms. the coexistence of old and new devices, as well

3
ADN

as 2G, 3G, 4G, and 5G. Significant increase in the Global telecom operators are making tremendous
network, service, and O&M complexity will lead to a efforts in making their networks autonomous and
significant increase in operational expenditure (OPEX). intelligent, finding ways to deploy AN over the
The average EBITDA margin of global operators is next decade in order to improve O&M efficiency
approaching 30%. Network automation technologies, and promote business growth. Their goal is to
such as SDN and network function virtualization deploy self-configuration, self-healing, and self-
(NFV), cannot overcome these challenges. Over the optimizing intelligent networks, develop efficient
next decade, the focus of network applications will O&M capabilities, and deliver new, high-quality
shift from consumers to industries. Providing agile, digital information communication services
on-demand, and personalized deterministic services featuring zero wait, zero touch, and zero trouble
and a differentiated experience for industry customers for end users, and public sector and enterprise
based on Connection+ will become the focus for customers.
industry competition. However, lack of business agility
on operator networks and slow service innovation As a new industry, AN has been embraced
(the average rollout time of new services is longer by all industry partners (including standards
than 12 months) create many obstacles for operators organizations, operators, and suppliers) since
competing with OTT or cloud service providers the concept was proposed and has witnessed
(who can roll out over 300 new services a year). tremendous efforts and innovation across all
As network infrastructure providers, operators also domains. Over the past two years, AN has
need to take on social responsibilities, such as carbon been incubated and is now in the planning and
emission reduction and security assurance. They development phase. The industry has reached
need an efficient and intelligent energy consumption consensus on the vision, target architecture, and
and security management mechanism based on the level standards of AN.
current architecture.

Delivering Zero-X experience

Zero-wait Zero-trouble Zero-touch

Service- Platform-
oriented Autonomous Networks oriented

Self-configuration Self-healing Self-optimizing

Deploying Self-X AN

Figure 1-1 AN objectives and vision

4
 ADN

Standard organizations, including TM Forum, are practicing AN based on four key elements
CCSA, 3GPP, and ETSI, have set up ad hoc topics (target architecture, level standards, effectiveness
on AN, initiated technical research and standards indicators, and operations practices) to develop
projects, and collaborated through the AN innovative scenario-specific applications. China
Coordination Meeting of M-SDO to preliminarily Mobile took the lead in setting the goal of
develop universal and domain-specific standards. achieving the Level 4 by 2025. It has evaluated
These standards cover five aspects: requirement the ANL for 31 provincial companies in China,
cases, reference architecture, level standards, and formulated targeted improvement strategies
key technologies, and key interfaces, focused and plans. China Telecom considers AN key to
on improving AN standards and laying a their cloudification and digital transformation
solid foundation for more efficient industry strategy and has announced the goal of reaching
collaboration and a thriving industry ecosystem. Level 4 in the short and medium terms. MTN
Multiple Standards Developing Organizations proposed the goal of achieving Level 4 by 2025
(M-SDOs) are collaborating with each other to to facilitate the Group's Ambition 2025 strategy.
formulate industry standards, while operators It has developed an AN framework featuring an
are speeding up AN deployment and practices. AN blueprint, agile development & operations
(DevOps) environment, high-performance
AN has become a must for operators looking network, and innovative high-value use cases
to implement digital transformation. They have (UCs).
set goals of achieving AN levels (ANLs), and

5
02
ADN Trends and
Suggestions

6
 ADN

2.1 Trend 1: In the Post-pandemic to a survey conducted by Deloitte from May to

Era, Home Office, Anywhere August 2020, 24% to 47% of people worked
Operations, and Online+Offline from home in 16 countries. This number
Education Are Becoming the New exceeded 40% in the UK, Belgium, Mexico,
Norms, Posing New Interaction and Ireland. Work from home has become one
Requirements for Home Networks of the key measures taken by enterprises and
individuals to cope with the pandemic.
Pandemic prevention measures (such as
home quarantine and social distancing) are Face-to-face teaching is changing to remote
moving a large number of operations and teaching or online+offline education when
economic activities from offline to online, schools in the basic education system are
profoundly changing business models. Non- shut down. The demand for online education
contact connection businesses and services are is increasing sharply, with a rapid growth
emerging rapidly, requiring home networks i n t he u s e r ba s e . The n u m be r of on li n e
to not only fulfill daily life and entertainment education users in China reached 342 million
purposes but also meet the requirements in 2020, with a year- on-year increase of
for remote office, learning, and production. 27.13%. In 2021, China invested over CNY
These requirements significantly increase the 400 billion in online education. Despite the
demand, usage duration, and quality standard fact that the pandemic is now under control,
for home networks. China still has up to 298 million online
education users, which is a slight decrease
Since the COVID-19 outbreak, a large number when compared with data from 2020. Online
of people have been working from home in fitness, language, and hobby courses have
many countries around the world. According matured and become popular choices.

7
ADN

These large-scale online activities would systems require higher video conference
be impossible without IT support in various quality, network security, and better and
industries, which is also provided remotely faster experience for certain applications.
due to the pandemic. Ser vices deployed
across distributed infrastructure are scheduled Second, anywhere operations are required
to preferentially meet digital and remote 24/7 to cope with the uncertainty of the
re q u i re m e n t s , e n s u r i n g a q u a l i t y o n l i n e pandemic. IT operations activities need to
experience for employees and customers. be machine- centered, instead of human-
According to Gartner, 40% of organizations centered, in order to reduce the workload of
will integrate virtual and real experiences operations personnel.
by 2023 to improve productivity and attract
more customers. Last but not least, large-scale online teaching
leads to a constantly high network load,
causing many problems, such as frame freezing,
audio asynchronization, and long-time image
Number of users in the
buffering. Live streaming for online courses
342 online education industry
million has high network requirements. A high-quality,
in China, in 2020
stable, and exclusive bandwidth of 20 Mbit/s to
50 Mbit/s is required for each course to ensure
a satisfactory video interaction experience.

By 2023, 40% of A future-oriented education experience and

organizations will integrate scenarios require virtual reality/augmented

reality (AR/VR) to deliver immersive,
40% virtual and real experience to
2023 holographic digital classes. Smart home is
improve productivity and the
becoming increasingly popular as people
customer coverage rate.
demand higher life quality. A large number
of smart Internet of Things (IoT) devices are
Network infrastructure is the cornerstone used at home, rapidly increasing the number
of all remote social activities and the key to of network connections. These devices include
continuous remote enterprise operations, intelligent surveillance, robot vacuum cleaners,
especially in the education and healthcare intelligent lighting, intelligent home appliances,
industries. Network infrastructure will become and intelligent locks. Different devices have
as indispensable as water and electricity. different network requirements. Some devices
require high bandwidth, some require scheduled
Diversified home network applications require control, and others require real-time control.
h i g h e r n e t w o r k q u a l i t y a n d p o s e m o re Users need a centralized home network-centric
requirements on home network O&M. application O&M platform to achieve smart
connections at home.
F i r s t , re m o t e o f f i c e re q u i re s p ro d u c t i o n
systems based on network infrastructure, such These trends and requirements create the
as an email system, internal communication following challenges for home networks:
system, video conference system, and Office
Automation (OA) system. These production

8
 ADN

voice, and game traffic, to prevent enterprise

information leakage or intranet attacks. In
addition, office applications, including Tencent
VooV Meeting, Zoom, WeCom, DingTalk, and
Challenge 1: Lack of Proactive Home Network Huawei desktop cloud, require application-based
O&M Leads to Inaccurate, Slow Locating of acceleration and SLA assurance capabilities to
Application-level Poor-QoE Issues, Affecting ensure optimal user experience.
Network Security

In work from home+office hybrid mode, video

conferencing has become the most effective
method for multi-region and multi-department
collaboration. The focus of broadband network
Challenge 2: No Operator Can Provide a
upgrade is often directed toward improving the
Quality Self-service Platform that Offers
download rate, instead of uplink bandwidth.
High-Quality Customer Services and
A low uplink rate will lead to many problems:
Assurance for a Large Number of Home
frame freezing in audio and video conferences,
Network Users
sound and image asynchronization, echoes, and
high cross-border access delay. These problems
In addition to remote office, anywhere operations,
affect the working environment and efficiency
online education, and smart home, self-employed
at home or office. A large uplink bandwidth is
people (for example, an owner of a small fruit store)
required to provide Service Level Agreement
are also home network users. They also require
(SLA) guarantee.
high-quality home networks for multi-connection
applications of goods management, cashiering,
During an online class, the teacher may need
monitoring, and online advertising, and flexible,
to interact with many students. Poor video
affordable, and reliable service packages. Customizing
quality on home networks affects the quality
a solution for each individual customer is a heavy
of teaching. Video quality needs to be detected
workload for operators. An optimal solution is
in real time to quickly identify deterioration,
bandwidth and routes need to be automatically to provide an e-commerce self-service mode for

adjusted, and acceleration technologies need to customers to subscribe to, change, maintain, and

be provided for teaching applications. optimize network service packages. Operators need
to develop a flexible and efficient e-commerce self-
A large number of terminals used by employees service platform for a large number of customers.
(including operations personnel) in the public Operators also need to identify typical service
sector and enterprises need to access the packages that can be recommended for different
Internet. This type of access has high security application types. Intelligent guidance needs to
requirements. If a terminal is infected with be provided in the entire self-service process to
viruses, the viruses may spread across the improve customer experience. These measures can
intranet through virtual private networks (VPNs). liberate IT operations personnel from the work of
For the purpose of information security, different communicating with customers and allow them to
channels are required for enterprise traffic focus on how to better integrate online and offline
and individual traffic, such as Internet access, experience.

9
ADN

intelligent methods are required. For example,

3D survey can be performed before home
network deployment to provide suggestions on
selecting an indoor network device deployment
Challenge 3: Home Wi-Fi Coverage Problems
solution, for example, fiber to the room (FTTR)
Lead to Frequent Complaints and Affect
all-optical gigabit network solution, or provide
User Experience
suggestions on device requirements and
deployment locations. After a home network is
Data of an operator shows that 60% of the
deployed, a mobile app can be used to detect
complaints from home broadband users are
the Wi-Fi signal strength, interference, and
caused by home conditions. Traditionally, home
network access rate at each location in the
network O&M only focuses on network access
house, and provide intelligent optimization
and does not focus on users' home conditions.
suggestions based on the user's house type and
However, users often have little knowledge
optimization cost. During network usage, the
or experience about networks. Any network
Wi-Fi signal coverage can be monitored to notify
problem may trigger user complaints. Wi-Fi
users of Wi-Fi coverage problems, and intelligent
coverage is a major problem that needs to be
guidance can be provided to help them solve the
resolved in order to reduce complaints and user
problems by themselves.
churn. Wi-Fi coverage problems include how to
reduce wall penetration loss, reduce co-channel
To overcome these challenges, Huawei proposes
interference, and optimize coverage for large
the following:
houses. It is impossible for O&M engineers to
manually solve all these problems due to various
house types and requirements. Automatic and

10
 ADN

1. CMOs of operators can seize the new opportunities brought by diversified

home services in the post-pandemic era, attach importance to guaranteeing
home network experience, proactively analyze poor-quality of experience (QoE)
customers, identify potential customers of new products, and provide high-quality
products that meet customer experience requirements. Meanwhile, anywhere
operations will bring new opportunities for cyber security authentication and
access, which require native security capabilities.

2. CTIOs of operators can focus on providing e-commerce online self-service

capabilities for new products to deliver a zero-wait, zero-touch, and zero-
trouble user experience.

3. CTOs of operators can pay more attention to home network Wi-Fi coverage
and experience assurance and develop self-service troubleshooting apps to extend
network O&M capabilities to homes and ensure optimal user experience.

2.2 Trend 2: Innovative Government cloud computing, innovative government and

and Enterprise Services Have Become enterprise services are opening up new business

the Main Growth Point of Operators' growth for operators. In the next few years,

Business, and the Profit Model Will the application of cloud-network synergy and

Change from Traffic Monetization to IoT in the industrial field will become the main

Service Monetization focus of operators' government and enterprise

services. Government and enterprise services are
With the maturity of technologies such as 5G and transforming from scenario-specific applications to

11
ADN

integrated solutions for vertical industry customers. In addition, the proportion of ICT+DT (DICT) cloud
The profit model of operators will change from service revenue to operators' government and
traffic monetization to service monetization. enterprise service revenue continues to increase.
Operators' cloud service revenue has doubled in
According to Gartner, by 2023, more than 60% of the past three years. For example, the proportion
enterprises will regard networks and clouds as critical of DICT revenue of China Mobile to the total
infrastructure that supports their digital strategies. revenue has exceeded 50%. In the next five years,
Driven by the convergence of 5G and cloud services, the application of cloud-network synergy and
operators' revenue for innovative services increased IoT in the industrial field, the new growth point
by more than 20% in 2021, accounting for 27% of of operators' business, will reach a compound
the total revenue. Innovative services are becoming annual growth rate (CAGR) of 19.1% and achieve
the second growth curve of operators' business. a global market scale of USD 600 billion by 2025.
Cloud transformation has become a consensus in
By 2023, over 60% of enterprises the telecom industry. In the next five years, global
will consider networks and cloud operators will increase their investment in cloud
60% transformation at a CAGR of 27%. By 2026, the
2023 as the key infrastructure for their
digital strategies. expenditure will reach USD 20.6 billion. In response,
operators' innovative services, including cloud,
By the end of 2021, 5G demonstration applications data center, video, and mobile payment services,
have been deployed in more than 20 industries are growing rapidly and gradually becoming the
around the world. 5G has broken through the second growth curve.
bottlenecks of network capacity and latency in
industrial applications, and has been verified as the
most appropriate technology in scenarios such as
600 20.6
steel turning/crown block in steel plants, remote billion (USD) billion (USD)

crane control at ports, and PLC cloudification for

3C manufacturing, creating significant value. For The global market Expenditure will
scale will reach USD 600 reach USD 20.6 billion
example, if the steel industry realizes automated billion by 2025. by 2026.
steel turning, the output value of one production
With the booming of government and enterprise
line can be increased by CNY 100 million in one year.
services, the boundaries of the telecom industry
Global operators have signed more than 10,000 5G
are becoming more blurred, and the competition
industry application contracts. GSMA predicts that
is diverse and complex. Operators need to
from 2022 to 2025, all industries will enter the in-
ensure scenario-specific network connections for
depth convergence phase; in 2025, 5G convergence
government and enterprise services and cooperate
applications will be popularized; and by 2030, the
with industry integrators to provide integrated
scale of the global vertical industry market enabled
solutions for vertical industry customers. Operators'
by 5G will reach nearly USD 700 billion.
profit model will change from traffic monetization
to service monetization. However, this is a gradual
The total market size in global development process that is faced with many
700 vertical industries empowered by challenges.
billion (USD)
operators' 5G networks is expected
to reach USD 700 billion by 2023.

12
 ADN

services, as well as the optimization of resource

configurations and the responsibility matrix to
reduce delivery and O&M costs.

Challenge 1: Existing Network O&M Mode To address the preceding challenges and
and Capabilities meet the requirements of the fast-growing
government and enterprise market , it is
During cloud transformation, there are various
recommended that both the network and
difficulties to overcome, such as a large number
operation systems be improved.
of access users and high requirements on
network reliability and SLA assurance. For 1. A ccelerate net work simplificat ion,
example, in the end to end (E2E) slicing scenario automation, and intelligence: Traditionally,
of a private network, the network reliability networks were deployed based on services.
must be higher than 99.999%, and the network With network cloudification, services are
latency must be under 5 ms. The network must carried by cloud-based networks in a unified
be able to automatically predict and optimize
manner. The traditional independent bearer
potential risks, diagnose and rectify poor-QoE
mode brings great complexity to E2E service
problems and faults, and quickly evolve to high-
control, increases joint commissioning and
level automation and intelligence.
interconnection costs, and cannot make full use
of resources. Therefore, network simplification
is required.

On the basis of fully decoupled architecture,

network resources are SDN- and NFV-
Challenge 2: Operation Capabilities and Cost
based to implement intelligent control of
of Transformation from ICT to OT
E2E networks. Cloud and network resources
support resource sharing, elastic scaling, and
During digital transformation, operators
on-demand allocation.
have much room for improvement in UC
development, value quantification, ecosystem
Intelligent technologies such as AI are
construction, E2E cost control, business model
introduced to transform the existing fault
collaboration, delivery and O&M capabilities,
O&M mode to an intelligent mode featuring
and internal organizations and processes. risk prediction, fault self-healing, and
Especially in the disruptive transformation from network self-optimizing. In this way, zero-
ICT to OT, operators are the most important trouble networks and zero-wait experience
part of the upstream and downstream industry are provided, meeting the core requirements
chain consisting of enterprises, integrators, of 2B customers for high reliability and
I S V s / I H V s , c l o u d s e r v i c e p ro v i d e r s , a n d differentiated SLA.
equipment vendors. In addition to coping
with fragmented applications, long processes, 2. Build a cloud-network integration operation
and many breakpoints in government and system: More flexible resource and capability
enterprise services, operators also need to be scheduling and collaboration are provided
responsible for the delivery and O&M of these through centralized display and unified

13
ADN

management of cloud and network and conduct integrated R&D, sales, and
resources and capabilities, enabling front- operations. In addition, they need to leverage
end services. Operators need to build a open IT capabilities such as AI, cloud
2B service provisioning and O&M system orchestration, and machine learning of cloud
that features cloud-network synergy across providers to accelerate the innovation and
the network management system (NMS), implementation of industry applications.
a core network, SIM card management,
c l o u d - e d g e c o l l a b o ra t i o n , a n d O & M ,

14
 ADN

2.3 Trend 3: AI Is Transforming Network communication is widely prevalent

Network Architecture from All-Cloud in our lives. People rely on mobile phones and
to AI-Native and Has Achieved Initial smart devices almost all the time. As more smart
Success devices such as smart cars and wearables access
the network, their influence on people will
Network architecture, which has transformed
increase sharply. The substantial transformation
from all-IP to all-cloud in the past two decades,
of human society depends on increasingly
is now evolving to AN based on AI-native. All-IP
intelligent, reliable, and flexible communication
solves the structural problem of traffic increase
networks. Therefore, communication networks
without revenue growth due to high network
are facing many challenges including large scale
OPEX, helping operators launch competitive
and complex services, O&M, and connections.
services to explore new revenue streams. All-
The best solution to these challenges is to deploy
cloud solves the problem of high costs stemming
from the management and maintenance of fully intelligent AN in order to simplify network

various network devices, and the structural O&M, improve network efficiency, implement

problem that operators could only increase network self-configuration, self-healing, and self-
network capacity by increasing the number optimization, and achieve network autonomy.
of devices on traditional telecom networks. In
the next decade, AI-native will be the core of According to GSMA's AI in Net work Use
network architecture. The control closed-loop Cases in China, "The convergence of AI and
and knowledge closed-loop based on network communications networks will inject new
digital twins will drive communication networks technological vitality into communications
to achieve full autonomy, thereby fundamentally networks and open unprecedented possibilities
solving the structural problems caused by a large and AI in Network is the key to success. This will
network scale and high complexity. truly promote the GSMA Intelligent Connectivity
vision and connect everyone and everything to a
better future."

The words in AI in Network Use Cases in China

are not just slogans. After several years of
exploration, remarkable achievements have been
made in network planning and construction,
network optimization and configuration, network
energy saving and efficiency improvement, as
well as in network maintenance and assurance.
TM Forum's AN whitepaper 4.0 contains the
latest AI network application success stories of
multiple operators.

15
ADN

China Mobile's intelligent operations of home broadband services: The accuracy of

proactive identification of poor-quality home broadband services is 95%, the user
experience is significantly improved by 83%, and the average fault locating duration is
reduced from 2.1 hours to 10 minutes.

China Mobile's intelligent optimization of wireless network coverage: On new 5G networks,

the average cell coverage is increased by 15.8%, the road coverage is increased by 91%, and
the outdoor download rate is increased by 30%.

China Unicom's intelligent operations of 5G bearer networks: The activation efficiency of

a single line is reduced from 45 minutes to 2 minutes, and the activation accuracy reaches
99.999% without manual intervention.

MTN's congestion suppression on mobile bearer networks: The traffic suppression model is
used to analyze real-time network congestion and provide guidance for accurate capacity
expansion. The average daily traffic of MTN South Africa is reduced by 28%.

Implementing network automation and intelligence processes. The evolution to AN enables human-
is a systematic project. Tremendous efforts have been out-of-the-loop network self-management,
made in this area and have achieved initial success. ultimately realizing always-on networks featuring
However, it still faces the following challenges: autonomy. With such a network architecture,
network O&M will transform from traditional
human-centered O&M to AIOps featuring
human-machine collaboration, and finally
evolve to intelligent O&M featuring human-
machine symbiosis. The evolution of the network

Challenge 1: Fully intelligent AN is an inevitable architecture and O&M mode will change O&M

trend. To smoothly evolve to such an innovative personnel's knowledge and work habits and

network architecture, especially Level 5, evolution promote the transformation of O&M personnel,

path selection is the key. There are two evolution organizations, and O&M processes.

paths: disruptive innovation and iteration. Operators

also need to choose between developing AN all on
their own and collaborating with industry partners.

Challenge 3: Collaboration among industry

organizations and alliances, including standards
organizations, is critical to the development
Challenge 2: The evolution of the network and large-scale application of AN. In the past,
architecture not only affects network architecture network automation standards were formulated
selection, but also changes the network O&M by different organizations separately without
mode and transforms organizations and coordination. AN involves more technical fields

16
 ADN

and requires stronger collaboration among loop and collaboration, network self-awareness,
industries and standards organizations in different network environment awareness and modeling,
fields. International and national standards self-fulfilling, self-healing, and self-optimizing
organizations, including TM Forum, ETSI, 3GPP, autonomous network, and network adaptive
IETF/IRTF, and CCSA, have carried out research learning and self-evolution. To make theoretical and
on AN in their own fields. How to remove barriers technological breakthroughs, collaboration is required
between the different organizations becomes between academia and industry in many fields,
especially critical. including software engineering, computing platform
and infrastructure, AI-based network cognition
system, system and control theories and applications,
human-machine interaction, information security,
and trustworthy network design.

Challenge 4: The control closed-loop and knowledge Network architecture transformation is a long-term
systematic project. The following are suggestions to
closed-loop based on network digital twins are the
address the preceding challenges:
basis of AN. Many key theories and technologies are
involved, including network digital twins modeling
and simulation, objective-driven automatic closed-

1. Evolving the architecture gradually, stably, and sustainably

Disruptive innovation involves self-development with operations support

system (OSS) as the core, whereas iteration involves the combination of
self-development and collaboration with network applications and network
infrastructure as the two cores. Path selection requires comprehensive
consideration of the product lifecycle from R&D to production and commercial
use, O&M mode transformation, and O&M capability inheritance. Evolution
of the network architecture shows that the iteration path is more suitable for

17
ADN

network architecture transformation. This path allows vendors and operators to

collaborate to achieve application-centric, data-driven, and intelligent AN. For
example, it allows vendors to implement single-domain autonomy and self-
closed-loop, while allowing operators to focus on new connection services,
customer experience, and E2E process streamlining.

2. Gradually establishing an intelligent domain O&M system based on

architecture evolution

Transforming the O&M mode requires organizational transformation. A

new O&M system should be gradually established from four aspects: O&M
organization structure, O&M positions, O&M scope, and O&M management
system and process. The new O&M system should be implemented by phase
based on the iterative evolution of the network architecture, gradually realizing
human-out-of-the-loop intelligent O&M featuring human-machine collaboration
and symbiosis.

3. Promoting the healthy development of AN through cross-industry

collaboration

Implementing AN is a systematic project that involves the entire industry chain, including
chips, devices, tools, integration, services, operations, and organizations. It is also a
transformation from partial autonomy to full autonomy that requires continuous
evolution. All industry organizations and alliances need to collaborate to define the
evolution pace, intergenerational characteristics, and implementation path.

4. Making breakthroughs in basic theories and key technologies through

cross-disciplinary collaboration

AN is also a new field that spans multiple disciplines, including communication

science, computer science, control theories, and complex systems. It requires
in-depth collaboration between academia and the industry for establishing
academic organizations (such as IEEE, ACM, and CCF) and interdisciplines
(such as interdisciplines based on computer soft ware) that meet AN
objectives, and for making comprehensive breakthroughs, from basic theories
to key technologies.

18
 ADN

It is recommended that the architecture design for AN comply with the following principles:

Six principles
——
Principle 1: Implementing hierarchical closed-loop autonomy
In the AN architecture, the resource operations layer is responsible
for real-time and near-real-time infrastructure resource supply and
assurance, the service operations layer is responsible for user service
enablement and user experience assurance, and the business operations
layer is responsible for business enablement, including offerings,
subscriptions, and billing.

Principle 2: Implementing continuous learning and evolution

Self-learning continuously optimizes knowledge and models to quickly
adapt to the dynamic environment, network status, device status, and
user behavior. In addition, exploratory intelligent awareness quickly
optimizes system parameters to ensure continuous system self-evolution.

Principle 3: Adopting the hybrid architecture driven by data and models

Most ICT infrastructure tasks are critical services. Incorrect decisions

made by a black-box system will lead to serious consequences. The
hybrid architecture driven by data and models improves the efficiency of
awareness and problem identification, and ensures reasonable decision-
making logic and predictable results.

Principle 4: Using human-machine symbiosis for task collaboration

To cope with uncertain and unpredictable tasks, such as handling
network congestion and frequent faults, the infrastructure dynamically
defines and initiates tasks and collaborates with humans to complete
them, instead of pre-defined tasks in the R&D and network design
phases. Ideally, backend personnel are responsible for specifying
objectives and policies and implementing intent-based translation, while
machines are responsible for breaking down objectives and directing
onsite personnel and machines to collaborate in executing tasks.

19
ADN

Principle 5: Using distributed intelligence for machine-machine

load balancing
To implement intelligent collaboration between network elements (NEs),
intelligent protocols, semantic interfaces, or intent-based collaboration
interfaces transfer status, semantics, or knowledge to each other in real
time or near-real time, and multiple systems and domains collaborate
to complete tasks. To implement intelligent collaboration between
autonomous domains (ADs), multi-domain, multi-vendor service
provisioning and network quality awareness and assurance are provided.

Principle 6: Enabling humans to take over systems

In case of a disaster, fault, or exception, humans can take over and
control machines to intervene and direct critical tasks.

2.4 Trend 4:Operators Are Accelerating applied AI to network O&M production and practices,
the Construction of AN, Transforming implementing proactive, preventive, and agile O&M.
from Passive to Proactive O&M, and
from Event-triggered Response to China Unicom built automation and intelligence

Proactive Prediction and Prevention capabilities based on the application-platform-

network layered structure, and developed a network
To achieve business agility and efficiency operations ecosystem featuring platform+application,
improvement, operators have developed automated intensive IT, and capability openness. China Unicom
and intelligent strategies and upgraded traditional also built a network AI capability center at the
siloed OSSs to platform+application systems. platform layer to apply AI capabilities to all network
operations scenarios.
China Mobile proposed a 2-3-4 architecture for
automation and intelligence capabilities, formulated China Telecom built a cloud-network foundation
a 2+5+N O&M architecture at the OSS layer, and and an AI-enabled platform. Vodafone performed

20
 ADN

intensive reconstruction at the OSS layer, integrated Uneven automation and intelligence capabilities at
OSSs of its operating companies (OpCos), and each layer make collaboration difficult and hinders
developed NaaS capabilities that enable efficient service innovation and business monetization.
operations and agile services. Telefónica developed Potential growths in digitalized vertical industries
a four-layer platform strategy and upgraded the AI and consumer life pose stricter and more diverse
capabilities of the 4th Platform to provide a distinct requirements on network bandwidth, latency, and
product experience. reliability, and call for a transformation from static
service provisioning to real-time adjustment. The
The operators are focused on transforming from automation and intelligence capabilities at the
passive to proactive O&M, and from event-triggered business operations layer, service operations layer, and
response to proactive prediction and prevention. resource operations layer lack centralized planning.
The quantifiable evaluation indicator system Complex interaction interface of each layer and
formulated by Vodafone and China Mobile is a differentiated automation and intelligence capabilities
typical example. The Zero-touch Operation (ZTO) make it difficult to achieve collaboration between
strategy proposed by Vodafone focuses on E2E Ticket layers, affecting service innovation efficiency and
Journey automation and sets goals on cognitive business closed-loop capability.
automation, configuration automation, outside
plant (OSP) digitalization and self-service, and AI Closed-loop autonomy has not been implemented
proactive prediction. Vodafone plans to achieve a for any domain, hindering the improvement of
fault automation rate of 80% and configuration network O&M management. The automation and
automation rate of 80%, and reduce 30% of passive intelligence level of each domain is still low, and this
fault ticket responses and 40% of OSP tickets by is not conducive for fault prevention and prediction,
2025. China Mobile has formulated effectiveness agile fault detection, root cause locating, and quick
indicators based on Zero-X and Self-X, including fault rectification. According to the data analysis on
service provisioning duration, service availability rate, 1800 typical O&M activities, 95% of processes and
customer self-service rate, configuration automation operation nodes still require manual intervention.
rate, fault automation rate, and poor-QoE self-
analysis rate. The external services of hyper-automation networks
lack evaluation standards. Standards organizations
are formulating standards on network automation
and intelligence to provide quantifiable evaluation
80% 30%
Configuration Passively
automation handled fault
Configuration
rate: 80%
Less
tickets: 30% less
solutions for network O&M capabilities of hyper-
automation networks. However, there is still a lack of
unified standards and evaluation methodologies for
the business values that AN can bring to end users. It
80%
Fault
automation 40% OSP tickets:
40% less is difficult to evaluate external service capabilities, and
Fault Less
rate: 80%
this impedes the development of hyper-automation
networks.
However, AN is a systematic engineering project. It
is faced with a series of challenges, including uneven To address these challenges, it is recommended that
capabilities at each layer, unrealized closed-loop telecom operators focus on the following points
autonomy in each domain, and lack of evaluation during AN construction:
standards for external services.

21
ADN

1. Making mid- and long-term plans by referring to best practices: CTIOs

should think three to five years ahead and learn from industry best practices of
TM Forum to define evolvable O&M architectures, goals, and implementation
paths. They can define service scenarios by business drives from top to bottom,
improve network operations from bottom to top, and evolve toward automation
and intelligence network based on service and network characteristics.

2. Managing network by layers based on industry standards: During

autonomous system construction, CTIOs can refer to the network automation
and intelligence standards formulated by standards organizations such as TM
Forum, CCSA, GSMA, ETSI and 3GPP to ensure collaborative innovation and
evolution of the business operations layer, service operations layer, and resource
operations layer. Layer-based network autonomy should be considered. CTIOs
can construct an autonomous system at each layer from bottom to top and
use intent application programming interfaces (APIs) between layers to simplify
interaction interfaces and achieve agile services.

3. Applying various technologies and building single-domain closed-loop

capabilities: To realize quick closed-loop troubleshooting in each domain,
operators and vendors should deepen technology collaboration, apply new
technologies such as AI, digital twins, and simulation in innovative practices.

4. Performing iterative evaluation for improvement and quantifying

effectiveness from multiple dimensions: Operators can refer to the practices of
standards organizations and leading operators, iteratively and quantitatively evaluate
the automation and intelligence capabilities of network O&M and the convenience
and flexibility of external network services, and iteratively improve network capability
weaknesses, ensuring orderly network transformation and quantifiable service
effectiveness.

22
 ADN

2.5 Trend 5: O&M Capabilities Are 80% and a configuration automation rate of 80%
Moving Close to Users and the O&M by 2025.

Mechanism for Handling Network

The process of delivering O&M capabilities close to
Faults Is Gradually Changing from
users and transforming the O&M mode is faced with
a Long-Process Ticket Transfer
the following challenges:
Mechanism to a Zero-Touch Man-
Machine Collaborative Processing
Mechanism

Cutting down unnecessary procedures to respond

to customers more quickly: Practice analysis shows
that an added procedure in the troubleshooting Challenge 1: Traditional O&M methods cannot
process increases the cost by USD 100 and increases adapt to the complexity of converged ICT
the customer waiting duration by 2 hours on networks. With the rapid development of the
average. With the development of digital intelligent network cloudification, virtualization, and service
capabilities, some operators have chosen to move the convergence, devices and service processes become

troubleshooting capability to the step closest to the more complex, alarms increase exponentially, and

fault to quickly realize closed-loop implementation, sub-health problems surges. Traditional methods

improving customer experience and E2E efficiency. of adding sites and personnel and relying on
expertise cannot meet the O&M requirements

Some operators are exploring ways to transform of complex services. The OPEX of operators has

the O&M mode in order to implement self- continued to increase in recent years. An industry

healing or one-hop closed-loop processing report shows that a doubled CAPEX investment of
an operator leads to three-fold increase in OPEX.
for events which affect services. For example,
China Mobile has proposed to build a fault
handling mode that integrates monitoring with
maintenance. A monitoring room is transformed
from "compressing alarms and generating tickets"
to "identifying, handling, and closing service-
affected events", that is, from "alarm-oriented Challenge 2: Traditional O&M processes

monitoring" to "event-oriented closed-loop". c a n n o t f u l ly u t i l i z e d i g i t a l i z e d O & M

This feature improves monitoring management capabilities. F o r e x a m p l e , a E u r o p e a n

efficiency, network service level, and customer operator uses a new automation engine to
experience. Vodafone has built an E2E automated implement E2E automated troubleshooting.
troubleshooting mechanism using Ticket Journey. However, the lack of a corresponding O&M
Based on capabilities such as cognitive intelligence organization and process transformation
and configuration automation, Vodafone plans makes production and implementation
to achieve an automatic fault closed-loop rate of challenging for its OpCos.

23
ADN

To re a l i z e h i g h - l e v e l A D N , w e n e e d t o develop new O&M mechanisms. We suggest

use cutting-edge technologies to improve the following:
capabilities and use advanced theories to

1. Finding a theoretical breakthrough point to ensure and improve the

availability and reliability of complex networks: Operators need to break the
linear increase relationship between network complexity and O&M costs.

2. Transforming O&M mechanisms, solving problems such as network faults

and user complaints with a team closest to end users: to achieve the goal
of 15150 (1 minute discovery, 5 minutes diagnosis, 15 minutes recovery, and
0 service impact).

2.6 Trend 6: Exploring O&M Values user satisfaction, 5% to 8% increase in user value,

from Edges. User Experience Is the and 0.5% to 1% reduction in user churn.

Primary O&M Goal and O&M KPI

To maximize data value and improve user
for Business Success
experience, operators are establishing dedicated
Using network and O&M data to improve user data analysis teams. Having realized the value
experience and effectively increase revenues: A of data, operators have constructed data lakes
Huawei practice result shows that network data over the past five years. However, it is impossible
utilization accelerates the user development speed to maximize data value by merely relying on
by 1.5x to 2x, and leads to 20% to 30% increase in traditional data mining and analysis team. More

24
 ADN

and more operators are setting up dedicated data

analysis teams. For example, LG U+ and XL Axiata
have established data departments. LG U+ has
set up a CDO position responsible for data service
integration and development. The core goals of Challenge 2: There is a lack of spatiotemporal

these operators are data-driven innovation, better methods for analyzing user experience. A

customer experience, and improved productivity. Chinese operator finds it hard to accurately
optimize user experience based on the

The process of exploring O&M values from edges traditional cell mechanism, which fails to identify

and improving user experience is faced with the the dynamic complaining time and locations.

following challenges: Operators need dynamic spatiotemporal analysis

methods to upgrade current measurement
methods and assurance mechanism.

Utilizing O&M data value to improve user

experience is one of the key points of high-level
AN. We provide the following recommendations:
Challenge 1: It is getting difficult to evaluate
user experience with indicators and rules.
According to a practice of a Latin American
operator, it is difficult to measure user experience
simply with network quality indicators. In some
areas with good network quality, user experience
is not as good as expected. In complex service
scenarios, user experience cannot be evaluated
simply by testing network speed.

1. Using AI technologies to evaluate user experience and providing support

for O&M value exploration: Network indicators and rules alone cannot
accurately or dynamically evaluate user experience. AI technologies can be used
to evaluate user experience in real time. Based on the practice and analysis of an
operator in Latin America, a network key quality indicator (KQI) and user experience
association AI model accurately identifies and optimizes poor-QoE areas, improving

25
ADN

the user-perceived rate by over 90% and releasing 71% more suppressed traffic.
AI technologies maximize the network data value that is not fully used and apply
network data analysis to develop the user base, increase user satisfaction, deliver
more value, and retain users, continuously improving intelligence on user- and
business-oriented decision-making.

2. Promoting spatiotemporal technology application and experience-

driven network optimization: The application of key capabilities needs to be
accelerated, including AI-evaluated user experience and Traffic Autonomous Zone
(TAZ). Driven by high business value scenarios requiring network optimization,
such as complaint handling, customer satisfaction improvement, and
performance benchmarking, operators should build a flexible and configurable
poor-QoE feature framework (which includes indicators highly relevant to service
experience and all-element data, such as events, time, and space data) based on
cross-domain data (experience data, CHRs, traffic statistics, dialing test data) to
provide optimal decision-making grounds for closed-loop network optimization.

2.7 Trend 7: IT Personnel Are involves 10,000 IT personnel in agile product

Transforming into Technical management and development, covering over

40% of employees. Telekom is upskilling personnel
Enablement Service Personnel. O&M
to adapt to future technical requirements. In
Knowledge Is Injected in Low-Code
Telekom's IT department, the proportion of
Mode Through Digital Intelligent
personnel skilled with key technologies, such as
Means for Machine Training and software/system engineers, DevOps engineers, data
Supervision scientists, architects, and agile experts, will increase
from 23% to 44% by 2024. (Transformation Case
Agility and DevOps create digital-intelligent
Study: Deutsche Telekom – 2021 )
positions: Since 2019, Deutsche Telekom has
shifted to a scaled agile framework (SAFe) that Low-code platforms lower the entry barriers:

26
 ADN

By 2024, more than 65% of applications will effectively applied to practical network O&M
be developed in low-code mode. 90.2% of to fulfill the 3-year goal of "reducing first-line
service personnel hope to quickly respond to support personnel's workload by 80% through
service changes in a low-code manner, 88.9% automation." The reason is that the cognitive
hope to improve the productivity of technology
automation team members (automation experts,
development and O&M management, and 73.4%
process analysts & engineers, and cognitive experts)
hope to explore agile innovation in a low-cost
are mainly IT personnel. Although this skill model
manner. (LowCode: China's Low Code & No Code
can quickly implement automation capabilities in
Industry Research Report in 2022 )
some scenarios, it is difficult to effectively control
%
maintenance services with complex scenarios
60

44% and discrete problems. Therefore, the model

45
cannot resolve actual problems in the production
30 23%
environment of each subnet.
15

0
Increase in the proportion of personnel skilled with key technologies

Operators are trying different exploration

methods: Over the past few years, Operators have
been cooperating with IT vendors to implement Challenge 2: The application effect of common
network O&M automation. A European operator low-code platforms in network O&M scenarios
is cooperating with an IT vendor. In later years, an is insufficient. Common low-code technologies
increasing number of operators have started to achieve agile development by means of balancing
cooperate with communications vendors. A case high flexibility and high encapsulation as well as
in point is a Chinese operator's cooperation with balancing communication-based orchestration
Huawei, which involves equipping service personnel and IT-based orchestration in common scenarios.
with scenario-based low-code technologies, long- However, these technologies do not make it
term O&M experience of the operator, and the easier for O&M personnel to manage services in
operator's understanding of Huawei technologies different scenarios, meet different network security
to make network O&M digital and intelligent. requirements, process different data formats, use
control instructions of different granularities, or
The transformation of O&M personnel faces the operate devices provided by different vendors in
following challenges: the same system. Therefore, the applications of
common low-code technologies are limited to
non-core scenarios, such as online management,
process orchestration, and third-party integration,
and are difficult to replicate.

Challenge 1: IT personnel dominat ion A highly skilled workforce is essential for digital
stifles in-depth digital-intelligent service transformation. We suggest increasing investment
reconstruction. A European operator cooperated in upskilling personnel to realize the evolution
with an IT vendor to build an E2E automatic toward high-level AN. We provide the following
O&M platform, which thus far has not been suggestions:

27
ADN

1. Encouraging and carrying out employee training and reskilling, and

driving service personnel to participate in the digital process.

2. Introducing a proper low-code platform to build development, data

governance, and analysis processes and methods: For example, China Mobile
Zhejiang and Huawei analyzed and abstracted fault cases of the cloud core network
spanning over a year, classified the faults into dozens of atomic troubleshooting
capabilities by means of systematic design, implemented high encapsulation of
network and O&M capabilities, and simplified the development process of upper-
layer service applications. Consequently, the service planning and rollout period was
shortened from three months to one month.

3. Combining organizational and process transformations with incentive

mechanisms to stimulate employees' creativity and encourage employees
to create value together with customers.

2.8 Trend 8: AI Foundation Models with the Existing Solutions, Such

Gradually Go Beyond Voice & Models Will Yield Breakthrough
Images into Vertical Fields. The Results
Telecom Industry Has the Huge
Amount of Data Required to Build Tech giants and AI research institutes around the
Foundation Models in Scenarios world are all researching pre-trained foundation
Like Energy Saving, Prediction, and models. Since Google released its foundation
Unknown Fault Handling. Compared model, BERT, in 2018, global tech giants have

28
 ADN

focused their efforts on the foundation model range of downstream tasks to meet different
technology. In the US, organizations such as application requirements. Foundation models have
OpenAI, Google, Microsoft, and Facebook have three main advantages:
developed foundation models with hundreds of
billions or trillions of parameters — prominent • Higher model development efficiency:
examples include GPT-3, Switch Transformer, and Foundation models are trained in advance.
MT-NLG. In China, we have many foundation When they are used in different scenarios,

models, including M6 from Alibaba, Wenxin from they only need to learn a little more to get

Baidu, Hunyuan from Tencent, and Pangu from themselves ready for the scenarios. This

Huawei. Foundation models have evolved from allows AI applications to be quickly deployed

NLP to CV, cross-modal, biological computing, and and applied, transforming manual work into

industry-specific models, and have enabled various automated assembly lines.

industries. Their applications in law, healthcare,

• Lower cost: Foundation models adopt a self-
education, and telecom have resulted in a positive
supervised learning method, so they are less
impact on industry development. In August 2021,
dependent on data labeling. To some extent,
more than 100 scholars, including prestigious AI
this method solves the problems of high cost
experts Percy Liang and Feifei Li, jointly published
and long periods that are characteristics of
the paper On the Opportunities and Risk of
manual labeling. Downstream developers can
Foundation Models, which systematically explains
get what they need simply by training small
the opportunities and challenges regarding pre-
samples, which greatly reduces development
trained foundation models. In the same year,
cost.
Stanford University established Center for Research
on Foundation Models (CRFM).
• Breakthrough: In order to improve model
precision, existing AI models t ypically
With the support of foundation models,
optimize the neural network structure, but
generalization is now possible with big data.
this approach hits a bottleneck. Foundation
Before the emergence of foundation models,
models, on the other hand, expand the data
AI development was by no means a "general-
and parameter scales, and will therefore
purpose" process. Any application in any scenario
improve model precision.
requires a group of technical experts to develop
dedicated AI models. If a new application is In future telecom networks, AI will be widely
required, development has to start from the very used in various service scenarios. Given that
beginning, which is as time-consuming and labor- telecom service scenarios have abundant data
consuming as manual work. This represents one samples, foundation model technology can be
of the biggest obstacles in implementing AI in leveraged to solve engineering problems, such
industries. Foundation models provide a new as AI generalization and model fragmentation,
solution to this long-standing problem. Built and the "pre-trained foundation models +
on huge amounts of data and innumerable downstream task fine-tuning" pattern can be
parameters, foundation models possess the basic introduced to accelerate the implementation of AI
capabilities that are required in a large class of and improve the effects of AI applications. After
scenarios. After initial development, the models being researched and practiced across industries,
can be fine-tuned and then applied to a wide foundation models have demonstrated their value

29
ADN

in some telecom scenarios: Taking core network to higher accuracy in identifying known faults,
as an example, the machine language foundation unknown faults can be automatically demarcated.
models that have been developed to learn features This signals a qualitative leap forward.
from massive amounts of data through a large
number of parameters, yielding more precise Foundation models have stringent requirements
models for core network fault demarcation and on computing resources and data volume. In
wider coverage of services. In terms of model view of these technical limitations, the telecom
development, secondary development of pre- industry still has a long way to go with regard
trained models shortens the time required to to the research and application of these models.
develop AI models and improves development The following suggestions may be helpful for
efficiency by 80%. In terms of application effects, developing foundation models in the telecom field:
horizontal demarcation across NEs and vertical
demarcation inside an NE take mere minutes to
pinpoint the smallest recoverable unit. In addition

1. Currently, the technical architecture of foundation models is applicable to a

limited range of scenarios. Considering the numerous service scenarios in the
telecom industry, it is unrealistic thus far to use a one-size-fits-all foundation
model to solve all tasks. A more feasible approach would be to start from a few
domains, for example, the core network. In the core network, which is mostly
cloudified now, it is easy to collect a large number of data samples from O&M.
Once the application effect is self-evident, the foundation models will be widely
adopted across the telecom field.

2. Foundation models need massive amounts of data, but it is difficult to collect

sufficient data samples from one operator network alone. To support the
innovation and research of foundation models, it is recommended that an
industry alliance be set up to organize all parties in building a unified data
sample library, which in turn will supply sufficient data samples to all parties
during innovation and research. This will also facilitate large-scale adoption of
telecom foundation models.

30
 ADN

3. Due to factors such as training costs, foundation models are fraught with heavy
investment and high risks. A wise move for the telecom industry would be
to "have big vendors pre-train models and operators apply the models". The
essence is to unite the efforts of different parties that have different focuses so
as to save on resources while also implementing more and better AI applications.

2.9 Trend 9: Cognitive Intelligence complex problems, cognitive intelligence can

Integrates Telecom Expertise with improve the explainability of intelligence.

the Advantages of Data-based Deep

Learning, and Is Expected to Become The development of cognitive intelligence

the Key Technology for Developing represents a new era, which will accelerate
the application of AI in the telecom industry.
the Next-generation Intelligent
There is a positive attitude in the industry
Network Operations System
with regard to the future of cognitive
Computational intelligence and cognitive intelligence. In June 2022, China's Artificial
intelligence are the different phases of AI. Intelligence Industr y Alliance (AIIA) and
Computational intelligence involves basic CAICT jointly released the White Paper on
logic operations and statistical analy sis. Knowledge Computing, pointing out that
Limited by regulated data logic rules, in the AI 2.0 era, data and knowledge are

computational intelligence cannot use the t wo most important basic elements,

expertise to perform in-depth, dynamic, and and various types of knowledge should be

heuristic reasoning. Cognitive intelligence fully used. The collaboration of knowledge

technology is regarded as the next phase and data will facilitate the innovation of

of intelligent development. It uses expertise AI algorithms and systems. According to a

to analyze, determine, and make decisions Deloitte survey, 40% of operators, media,
and technology executives say they have
based on its understanding of data logic.
benefited from cognitive intelligence, and
Cognitive intelligence effectively integrates
25% have invested more than USD 10
data and knowledge, and further promotes
million in it. More than 75% expect cognitive
AI and system innovation. Besides solving

31
ADN

computing to "substantially change" their and identifies diagnosis results based on

companies over the next three years. In knowledge, performing high-precision fault
addition, Gartner released the Emerging diagnosis in complex scenarios.
Technologies and Trends Impact Radar 2021,
which listed Advanced Virtual Assistants • Cognitive intelligence enables knowledge
( with cognitive intelligence at its core) as evolution in the telecom industry. Take the
one of the 23 most influential technologies. core network as an example. Cognitive
Some telecom operators are attempting intelligence is used to obtain knowledge
to incorporate knowledge into the overall about fault propagation in machine data
digital intelligent network solution based on and automatically confirm the knowledge
the cognitive intelligence technology, build using technologies such as abduction,
a Knowledge as a Service (KaaS) framework, thereby implementing self-learning and self-
and take the lead in integrating knowledge evolution.
into fault diagnosis.
Evolution towards ADN Level 5 means that
In the telecom industry, services are heavily people are less in volved in the net work
reliant on expertise and require secure and lifecycle. Cognitive intelligence technology
reliable AI applications. Based on current uses human-like decision-making methods
research, cognitive intelligence is expected to t o m a k e t h e n e t w o r k m o re i n t e l l i g e n t ,
inject new impetus to network intelligence. supporting the evolution towards Level 5.
In fault O&M scenarios, this technology
• As a national infrastructure, telecommunication increases the fault demarcation accuracy to
relies upon secure operations and explainable more than 95% and makes fault demarcation
inference that underpins major operations. results highly explainable. Recovery
Cognitive intelligence applies knowledge commands can be automatically generated
deduction to inference. Take fault diagnosis based on knowledge to implement automatic
as an example. Besides fault demarcation, n e t w o r k r e c o v e r y. I n c a r r i e r n e t w o r k
knowledge inference+machine learning can O&M scenarios, cognitive intelligence
visualize the propagation and causal process of technology can be used to automate fault
abnormal events, which improves explainability. O&M and implement self-learning of
fault knowledge. As a next-generation AI
• Cognitive intelligence makes intelligent technology, cognitive intelligence is about
applications perform better in complex generalization and explainability. Being a key
scenarios. Cross-domain fault demarcation technology for ADN, cognitive intelligence
and locating in telecom networks involve is still in the innovation and exploration
various types of NEs, data sources, and phase. To accelerate its application in the
multi-layer deployment resources (NEs/PaaS/ telecom industry, we provide the following
IaaS). Traditional AI cannot apply expertise suggestions:
in complex scenarios, making challenges in
precise demarcation difficult. The cognitive
intelligence technology integrates the
relationships between networking and service
modules into fault diagnosis, and analyzes

32
 ADN

1. Applying cognitive intelligence in the O&M domain first in view of the

complex O&M scenarios and massive expertise requirements: Based on the
idea of integrating knowledge with data, cognitive intelligence technology can be
used to implement automated, accurate fault O&M, and improve explainability.

2. Starting from the usage of virtual assistants: Before ADN Level 5, man-machine
symbiosis is still required where virtual assistants based on cognitive intelligence can
collaborate with people and gradually participate in network decision-making and
execution to implement self-learning and self-evolution. The autonomy and self-
healing of ADN Level 5 in O&M scenarios can be achieved.

3. Establishing standards: The telecom industry standards for knowledge should be

established to facilitate knowledge sharing, interaction, and development of cognitive
capability by all stakeholders in the industry.

2.10 Trend 10: Digital Twins Become for three consecutive years from 2017 to
a Crucial Intelligent and Automation 2019. Digital twins are expected to reach the
Technology for the Full Network Plateau of Productivity in the Gartner Hype
Lifecycle Cycle over the next five years.

Since Dr. Michael Grieves first proposed the

In the field of smart manufacturing, digital
concept of digital twins in 2002, it has been
successfully applied in various industries, model design technologies are used to map

including city construction, aerospace, and the attributes of physical devices to a virtual
smart manufacturing. It was elected one of space, generating digital images that can be
Gartner's top 10 strategic technology trends disassembled, replicated, transferred, modified,

33
ADN

deleted, and repetitively operated. In the that monitors everything from bus stations to
three-dimensional digital virtual space, the buildings. Cityzenith has deployed a 5D Smart
sizes and assembly relations of components City to digitalize infrastructure development and
and products can be quickly and conveniently city management throughout the entire lifecycle.
modified, significantly reducing the number of
physical prototypes and reducing the time and The role of network infrastructure is becoming
cost of making the prototypes. In addition, increasingly important in a developing global
digital twins collect a limited amount of key economy. Home quarantine, as a most effective

performance indicator (KPI) data from sensors, measure against the COVID-19 pandemic,

and use historical data patterns and machine has increased the demand for Internet access.

learning to predict KPIs that cannot be directly The net work load and scale continue to

m e a s u re d . T h i s m e t h o d c a n b e u s e d t o increase, and the number of connections is

evaluate the current status, diagnose problems, growing exponentially and is expected to
reach 200 billion by 2030. Meanwhile, cloud
and predict future trends. The analysis results
and virtualization technologies are increasing
can be used to simulate all possible scenarios
n e t w o r k c o m p l e x i t y. W e a r e f a c e d t h e
and facilitate decision-making. After being
following challenges:
developed for over more than a decade,
digital twins have been successfully applied
1. Evaluating the impacts of network changes
in urban construction, satellite networks, and
becomes a new challenge due to the lack of
production environments.
simulation and verification methods.

In the field of smart city, digital twins improve

2. No effective method is available to evaluate
the quality and standard of city planning,
the risks and potential threats, affecting
facilitate city design and construction, and assist
network stability.
city management and operation, improving
the urban environment. Take Xiong'an New
3. It takes longer to develop innovative network
Area, a state-level new area in the Baoding
technologies and services.
area of Hebei, China, as an example. It has
set the goal of evolving towards a world- Digital twins can create virtual images of
leading digital city. A digital twins city includes physical network infrastructure to build a
five aspects: physical city, virtual city, big data, digital t win net work with the same NEs,
virtual-reality interaction, and intelligent services. topologies, services, and data as those
Sensors are deployed on urban construction on the physical network. The digital twin
devices to detect and monitor the city's running network is a test bed for verifying network
status. A digital twins model is developed to c o n f i g u ra t i o n s a n d n e w t e c h n o lo g i e s . I t
simulate the entire physical city. Running data can significantly reduce live-network risks
is collected to optimize the digital twins. The a n d p re v e n t f a u l t s c a u s e d b y i n c o r re c t
virtual-reality interaction of the digital twins c o n f i g u ra t i o n s . I t i s a l s o ke y t o va r i o u s
can provide intelligent services, such as city scenarios, such as holographic visualization
planning and design, and municipal planning o f n e t w o r k t ra f f i c a n d f u l l - l i fe c y c l e N E
optimization. Singapore has collaborated with management. Real-time interaction between
Dassault Systèmes to build a digital twins city the physical network and twin network can

34
 ADN

lower the trial-and- error cost , facilitate 2. M o d e l d e f i n i t i o n : L a y e r e d n e t w o r k

i n t e l l i g e n t d e c i s i o n s , a n d i m p ro v e s t h e abstraction is implemented to study and
efficiency of innovation, achieving simplified define layered abstract models oriented
and intelligent O&M. to network twins. The models are defined
based on data for various net work
D i g i t a l v i r t u a l t w i n s c a n b e c re a t e d fo r applications.
physical network entities in order to build
a system that can interact with and map 3. Twin mirroring: Network twins are used to
physical network entities in real time. The map physical network entities on a high-
core elements of the digital twin network fidelity page in visualized mode in order
are data awareness, model definition, twin to form a twin mirror for the twin network
mirroring, and virtual-reality interaction. and physical network.

1. Data awareness: High-performance approximate 4. V i r t u a l - r e a l i t y i n t e r a c t i o n : V a r i o u s

network measurement is studied to achieve technologies, including big data analysis, AI
approximately error-free measurement. algorithms, and cognitive intelligence, are
Data at the physical network layer, including used to implement full-lifecycle analysis,
physical entity data, space data, resource di a gn os i s , s i m u la t i on , a n d con t rol of
data, protocols, interfaces, routes, signaling, physical networks, achieving closed-loop
processes, performance, alarms, logs, and control of the digital twin network and
status, is collected and stored to the data physical network.
warehouse in real time or non-real time.
The data is then used to build and empower
network twins.

35
03
Huawei ADN
Solution

36
 ADN

The vision and mission of Huawei is to bring digital to every person, home, and organization, and build
a fully connected, intelligent world. Building a fully connected, intelligent world is a process of helping
various industries to implement digital transformation. Huawei is committed to continuous innovation
in cloud services, AI, networks, and low-carbon development to accelerate digital development, spark a
new revolution for human civilization, and stimulate industry innovation, industry upgrade, and social
development.

AI

Low-carbon development benefiting

Ubiquitous cloud services Pervasive AI ADN
from digital technologies

Huawei ADN solution is designed to embrace operators and enterprises, delivering a zero-wait,
the trend in the AN industr y. As one of zero-touch, and zero-trouble experience to both
Huawei's four initiatives for continuous digital individual consumers, and public sector and
innovation, ADN aims to leverage connectivity enterprise customers.
and intelligence to deploy a self-fulfilling, self-
healing, and self- optimizing autonomous According to Huawei's Intelligent World 2030
network . ADN implements single- domain report, a seamless intelligent network that covers
autonomy and cross-domain collaboration to land, sea, and air is the future. The network
develop self-configuration, self-healing, and will provide ubiquitous cloud-edge-device
self-optimizing network O&M capabilities with convergence services in all production scenarios,

37
ADN

achieving fully autonomous communications advent of a more intelligent world. ADN is a core
networks. Such communications networks will strategy of Huawei Communications Network
continue to play a leading role in promoting the 2030.
development of the world and accelerate the

3.1 Architecture and Features of Huawei ADN Solution

The telecom industry needs a clear service architecture as a reference and the consensus to carry out
production practices in order to achieve the goal of reaching high ANLs. Based on this architecture,
operators can systematically evaluate and analyze the architecture of current OSSs, integrated NMSs,
vendor NMSs/controllers, and network devices from top to bottom, and formulate a feasible evolution
roadmap that meets their own requirements. Based on the suggestions for the service architecture
provided by TM Forum in IG1218 and various innovative practices jointly carried out with global
operators, Huawei proposes the following ADN service architecture.

ADN architecture
Carrier BSS

Cross-domain
service units
intelligence
Service

Business enablement
Design Studio (automation
and intelligent cognition Customer experience
Carrier OSS
application design)
Intelligent O&M

Cross-domain collaboration

·
Network management ·
and control units Network management and control units
intelligence

(Domain X) (Domain Y)
Network

Management Control Analysis … Management Control Analysis

EMS/NMS
Network AI Engine Network AI Engine

Single-domain Single-domain
autonomy autonomy
intelligence

AI inference framework
Network
NE

infrastructure …
Real-time awareness
Wireless Access Transport IP Core

Figure 3-1 ADN service architecture

38
 ADN

Intelligent network infrastructure is fundamental unit also acts as an asset center for operators'
to high-level AN. Network devices need to AI assets. It manages AI models and network
have more real-time sensing components knowledge developed and trained by operators
and AI inference capabilities to detect the during planning, construction, maintenance, and
changes of resources, services, and surrounding optimization. The AI models and knowledge can
environments, and the capabilities for intelligent be fully shared and reused to reduce repetitive
analysis, decision-making, and execution. development and training.

A network management and control unit Features of the ADN Solution

integrates network management, control, and
Native Intelligence
analysis modules to automatically translate
upper-level intents of services and applications As the foundation of ADN, network infrastructure
into network behavior, realizing single-domain plays an indispensable role. For AN to advance
autonomous closed-loop and continuous SLA toward higher levels, the network infrastructure
assurance. Meanwhile, based on the built-in needs to provide more powerful capabilities,
network AI engine, it continuously conducts AI including real-time awareness, local reference,
training and knowledge extraction to constantly and edge computing. As a thriving technology,
strengthen AI models and enrich network AI empowers the infrastructure layer with new
experience, making the network more intelligent. capabilities. NE AI uses distributed computing
power to implement local inference based on
Cross-domain service units provide three a large amount of KPI data generated by NEs.
capabilities (intelligence business innovation, This enables quick risk awareness and analysis,
customer experience improvement, and service while reducing the amount of data sent to the
and network O&M assurance), an application network layer and the amount of computing
design and development platform, and cloud power required in the management system. NE
services. This platform focuses on O&M and AI also implements AI-based edge inference and
business process streamlining, and flexible service NE-level quick fault self-healing and decision-
orchestration based on network characteristics. making.
It can be used to quickly and iteratively develop
new business models, O&M processes, service For example, Huawei's innovative IP service flow
applications, and business products and services. detection protocols, such as In-situ Flow Information
This is the key for operators to achieve business Telemetry (IFIT), enable EMSs to quickly and
agility and improve the skills of new O&M automatically locate faults in complex IP network
personnel and business design personnel. topologies when service poor-QoE problems or faults
occur. IPv6+ greatly simplifies service provisioning
The network AI unit, a basic platform for processes and improves the automation level of
network AI design and development, performs service provisioning. AEC boards on access network
continuous AI training and knowledge extraction optical line terminations (OLTs) can analyze
for various types of telecom network data to application KPIs or KQIs to identify poor-QoE users.
generate AI models and network knowledge. The optical iris technology of DQ-ODN can visualize
The AI models and network knowledge can last-mile dumb fiber resources. Huawei's optical
be injected into the other three units to make transmission products can identify co-cable fibers to
the network more intelligent. The network AI prevent SRLG problems.

39
ADN

Single-Domain Autonomy For example, intelligent energy saving for

wireless base stations requires dynamic and
The network infrastructure layer and the
precise adjustment of many device parameters
network management and control layer are
b y a n e t w o r k m a n a g e m e n t a n d c o n t ro l
coupled. Different network domains have
system at the base stations. Huawei's network
different characteristics. The increasing network
complexity and improved capabilities in terms management and control systems can predict

of net work infrastructure make net work cell traffic based on real-time network data and
management and control more complex . quickly adjust energy saving policies to realize
In particular, cross-layer and multi-domain energy saving while keeping KPIs stable.
management becomes challenging. Autonomy
must be achieved for each network domain Open Collaboration
before cross-domain management can be
Most telecom services are dependent on single-
realized through cross-domain orchestration.
domain autonomy and open collaboration.
Different ADs provide standard northbound
interfaces (NBIs) for the upper-layer OSS or The upper-layer OSS and service layer focus

service orchestration system. The upper-layer on service policies and innovation, and the
OSS or service orchestration system invokes and lower-layer network management and control
orchestrates the capabilities of each domain to layer and infrastructure layer focus on service
provide services across multiple domains. Devices implementation and technology innovation.
at the network infrastructure layer introduce Standard intent APIs are developed for the
many new functions. Standard interfaces cannot network management and control system in
be quickly developed for these new functions. each AD to completely decouple the network
Coupling devices at the infrastructure layer and management and control layer from the service
the network management and control system
layer and OSS layer. Services and OSSs invoke
of the device vendors can speed up and improve
these APIs to implement cross-domain service
single-domain autonomy.
management, including service provisioning and
fault correlation analysis.

Take service provisioning as an example. An OSS

uses open APIs to deliver service intents, such
as source and sink nodes and service SLAs, to
the network management and control system
in each domain. The network management and
control system in each domain independently
plans and prepares functions and parameters
required for provisioning services in the domain
and deliver configurations to devices in the
domain, including device addresses, tunnel
addresses, tunnel types, and tunnel parameters.
Finally, multiple domains collaborate to quickly
configure services.

40
 ADN

3.2 Panorama of Huawei ADN Solution

Since Huawei first proposed the concept of ADN at UBBF 2018, it has developed a series of innovative
ADN schemes for various domains, including wireless, core, access, transport, IP networks, data centers,
and enterprise campuses. Adhering to the concept of Intelligence for ICT, Huawei is committed to
systematically applying AI technologies to ICT infrastructure in order to overcome key challenges and
fulfill the vision of building an intelligent world.

IntelligentServiceEngine

User Intelligent Scenario-

Virtual
experience business based O&M
employee
AD enabler AD

IntelligentCore

MAE-CN
IntelligentRAN MEC
Individuals

…
SPC SVC
MAE MEC

MetaAAU
IntelligentWAN

Public …
NCE-IP cloud
IntelligentFAN
IPv6+ IPv6+ Huawei Cloud
Homes

Access Metro Core

NCE -FAN Router Router Router
FTTR
DQ ODN
OLT

NCE-T
Enterprise
IntelligentCampusNetwork DCs
Enterprises

IPv6+ Access Metro Core Computing

OXC OXC OXC
Wi-Fi 5G Ethernet POL IntelligentStorage
CPE
… IntelligentFabric
IntelligentOTN

Figure 3-2 Products and panorama of the ADN solution

3.2.1 IntelligentRAN — Wireless ADN

As the Huawei ADN solution for wireless networks, IntelligentRAN will enable wireless networks to

41
ADN

evolve from automatic O&M to intelligent networks. IntelligentRAN focuses on intelligent functions on
the RAN side. It uses iMaster Mobile Automation Engine (MAE) to implement single-domain single-
vendor basic O&M in the wireless domain. In addition, it features the Mobile Intelligent Engine (MIE),
providing intelligent UCs in cooperation with NEs. Intelligent atomic capabilities are opened to the
cloud through an intent interface.

OSS

IntelligentRAN

iMaster MAE Openness Intent-driven

MIE Data & Platform App

Northbound
interconnection Network Digital Twins Experience iApp

View Domain: topology, config., performance, etc.

Service iApp
FCAPS
Network: coverage, energy consumption,
experience, etc.
Entity
Environment
& users
Energy efficiency iApp
NE: specifications, config.,
algorithms, etc.
NE access
O&M iApp
Data engine AI engine
…

Data Model Decision-making

coordination coordination coordination

Base station Data modeling

Target Target Target Target
…
Data L3 smart grid 1 2 3 4
collection
O&M management
L2 scheduling dictionary iRRM algorithm
Data
On-demand resource
sensing L1 channel graph iRTT algorithm
orchestration

Transmission module
self-configuration AI framework/algorithm component library

Figure 3-3 IntelligentRAN

Highlights diversified, and net work complexit y will

continue to increase. The following structural
With the continuous evolution of mobile
challenges of mobile networks will become more
networks from 2G, 3G, and 4G to 5G, 5.5G, and
prominent.
6G, mobile network capabilities will be further
upgraded and services will be increasingly 1. Challenge 1: how to implement simplified

42
 ADN

O&M with increasing network complexity • Intelligent service operations: Based on

differentiated service SLA requirements,
2. Challenge 2: how to cope with 100-fold user-level dynamic simulation is used to
traffic growth with energy consumption implement precise network planning based
slightly increased on coverage, rate, and latency, enabling
fast service deployment and provisioning. In
3. Challenge 3: how to ensure optimal service addition, based on prediction capabilities,
experience while providing diversified services real-time dynamic resource scheduling is
implemented to ensure service experience
In response to these challenges, IntelligentRAN and implement service-oriented network
integrates intelligent capabilities into wireless adjustment.
net work ser vices, experience, O&M, and
energy saving to build a wireless ADN network Key UCs
that features intelligent and simplified O&M,
1. Intelligent and simplified O&M
intelligent network optimization, and intelligent
service operations.
• P re d i c t i o n - b a s e d f a u l t m a n a g e m e n t ,
transforming troubleshooting from passive and
• Intelligent and simplified O&M: Based
responsive O&M to proactive troubleshooting:
on key platforms and technologies such
ToC and ToB services carried on mobile
as the expert knowledge base, prediction
networks are required to be always online.
algorithm, and neural network, intelligent
Traditional responsive O&M based on post-
alarm management in the wireless domain is
event work orders cannot meet these
provided to accurately identify alarms, quickly
requirements. iMaster MAE accurately
locate faults, and predict and prevent faults.
identifies alarms and locates faults by
This feature helps operators transform from
enhancing fault awareness capabilities.
responsive O&M to predictive and preventive
In addition, iMaster MAE predicts and
O&M, achieving "zero" network faults.
analyzes software and hardware faults and
performance deterioration (such as high
• Intelligent network optimization: Intelligent
temperature of boards, optical module
NEs are introduced for intelligent resource
faults, and short power backup duration)
s c h e d u l i n g i n o rd e r t o o p t i m i z e u s e r
on the network based on long- and short-
experience and capacity and maximize
period awareness data obtained through
spectral efficiency in the multi-band and
c o l l a b o ra t i o n b e t w e e n t h e N M S a n d
multi-site heterogeneous net work . In NEs. This facilitates fault rectification and
addition, the net work-level intelligent eliminates risks, achieving "zero" network
engine is used to implement intelligent faults and improving performance reliability.
collaboration based on multiple intents
and objectives, helping operators transform 2. Intelligent network optimization
their goals from optimal performance to
optimal performance+energy consumption, • Intelligent selection of multi-band carriers
maximizing network energy saving while based on smart grids, achieving optimal
ensuring a stable network performance. network performance: A wireless network

43
ADN

typically uses multiple frequency bands, This maximizes the network energy saving
each of which has different characteristics. effect while ensuring a stable network
Coordination between frequency bands is performance, and strikes an optimal balance
critical to improving the spectral efficiency of between energy consumption and KPIs. It
the entire network. In complex multi-band enables the transition from performance-
networks, intelligent schemes are used to oriented networks to networks that balance
improve multi-band coordination efficiency performance and energy consumption.
and combine the advantages of different
frequency bands to achieve optimal network 3. Intelligent service operations
performance.
• SLA-oriented precise network evaluation
• Multi-intent-based multi- dimensional and planning, facilitating precise and fast
collaborative energy saving, optimizing service provisioning: There are a variety
both performance and energy saving: Low- of 5GtoB services. The services have high
carbon emissions have become a common SLA requirements and the application
concern of the communications industry
environments are complex. The traditional
and the global society. Energy efficiency
mode of manual network planning based
has become an important indicator for
on expert experience cannot meet the
mobile communications networks. Intelligent
deterministic network planning requirements
a l g o r i t h m s a n d m o d e l s a re t h e re fo re
of thousands of industries. In a scenario
introduced to determine optimal policies
involving public network for dedicated use,
for saving energy dynamically, multi-
the network coverage and data rate of
dimensionally, and coordinatively based
given areas are accurately evaluated, greatly
on the traffic and environment, achieving
reducing the cost of onsite measurement
network scenario auto-adaptation, one policy
and evaluation. In industry-specific network
for one site, and multi-network coordination.
scenarios, industry profiles, environment
modeling, user-level simulation evaluation,
and SLA-based precise planning are used to
meet SLA requirements of various services on
the live network, improving network planning
efficiency.

• Prediction-based slice SLA assurance,

enabling service-oriented network changes:
Traditionally, resources are reserved for
slice SLA assurance. To ensure services run
stably, largely redundant resources are
reserved, resulting in resource idleness. The
prediction capability is therefore introduced
to implement adaptive slice SLA assurance,
improving resource utilization while ensuring
slice SLA.

44
 ADN

3.2.2 IntelligentCore — Core Network intelligent analysis, model training, AI inference,

ADN and intent insight, are introduced to provide
automation and intelligence capabilities at
Adhering to the cloud-based O&M transformation
different network layers (including the telecom
concept "high network stability, intelligent and
cloud base, 5GC, and SVC at the NE layer, and
simplified O&M, and optimal user experience",
iMaster MAE-CN at the management and
Huawei core network ADN solution integrates
control layer). In this way, this solution helps
and innovates many core network products and
operators build a core network that features full-
professional service tools to make network-
process automation, automatic self-healing, and
wide data assets visualized, manageable, and
self-optimization.
traceable. AI technologies, such as data native,

OSS

IntelligentCore Intelligent network management

and orchestration
iMaster MAE-CN
Network Network Network
intent cognition control

Network Network AI
simulation digital twin framework

Functional plane, on-demand expansion

Intelligent AI
services Digital twin framework
API API
gateway gateway
Device Session
management
Mobility
management Authentication Application

Hyper-distributed UPFs and computing services

Intelligent AI
services Digital twin framework
…
Application Forwarding Computing
awareness services services

Figure 3-4 IntelligentCore

45
ADN

Highlights optimization: Native AI is used to transform

from network quality-centric management
Catering to the requirements of cloud-based to personalized user experience-centric
O&M transformation and new 5G services, management (service experience standard,
t h e H u a w e i c o re n e t w o r k A D N s o l u t i o n experience awareness, and experience closed-
builds capabilities in network maintenance, loop), and provide enterprise-level service
operation delivery, and experience closed- SLA assurance for customers in different
loop management, helping operators achieve
industries, achieving experience monetization.
intelligent data transformation.

Key UCs
• High stability of always-online services: Based
on expert experience, knowledge graph, AI self-
1. Fault Scenario — High-Stability Network
learning, and digital simulation, single-domain
fault maintenance of telecom cloud and core The core network is crucial for the entire
network NEs is implemented in a closed-loop
network. It bears services for a large number
manner. NE/module-level fault management
of users, and its fault has a wide impact scope.
capabilities are built to implement sub-health
Therefore, high stability is always a priority for
prediction, precise fault demarcation, and quick
the core network. Huawei core network ADN
self-healing. This ensures that always-online
solution introduces intelligent methods to
communication services can be provided for
provide routine DR drill, health check, prediction,
consumers and enterprise users, and enables
and diagnosis and troubleshooting capabilities,
operators to transform their O&M mode of
ensuring high stability and reliability of the core
cloud-based infrastructure from passive response
network.
to proactive drilling, detection, prevention, and
treatment.
• Health check : The solution evaluates
network reliability, ensures the availability of key
• E 2 E n e t w o r k o p e ra t i o n a u t o m a t i o n :
Intelligent and simplified delivery capabilities resources, and prevents emergency faults.

are built for routine operation scenarios, such

• Routine drill: The entire process is visualized and
as upgrade, capacity expansion, migration,
and testing. Based on the digital twin the DR drill efficiency is high.

technology and cloud-based Continuous

• Expert diagnosis: The solution provides fast
Integration Continuous Delivery (CICD) tool
chain, the solution exponentially reduces the fault detection and accurate demarcation and

complexity of routine operations on cloud- locating capabilities. Faults are demarcated to the

based networks. Driven by user intent, the minimum risk operation unit, facilitating service

solution streamlines the delivery process, recovery.

achieves automatic delivery throughout
the entire process based on the pipeline, • Health prediction: The health management

and ensures lossless and secure change mode is transformed from passive handling
operations. to proactive prevention, and subhealth
issues can be prevented without human
• R e f i n e d a n d p e r s o n a l i z e d e x p e r i e n c e intervention.

46
 ADN

2. Network Change Scenario — Intelligent 3. Optimization Scenario — Optimal User

and Simplified Operations Experience

Traditionally, network changes (such as upgrade User experience management faces new
and capacity expansion) are performed manually. challenges in different phases of 5G commercial
The operation procedure is complex , and use. Good user experience is the basis for user
requires a large amount of manual intervention, satisfaction. Huawei core network ADN solution
leading to low efficiency and high possibility of transforms the goal from being network-centric
upgrade risks. Based on the advanced DevOps into being user experience-centric and value-
concept in the industry, Huawei core network centric, improving user experience.
ADN solution implements efficient and secure
change operations. • Experience optimization: To improve user
experience for key services such as VoNR and
• Automation: The E2E tool chain eliminates ViNR, native AI is used and the tool chain
the breakpoints of manual operations, reduces is optimized. This ensures quick call setup,
manual intervention, and improves the delivery high voice quality, and zero call drops during
efficiency in all scenarios, including upgrade, communication, reducing the subscriber
capacity expansion, and cutover. churn rate.

• Lossless change: The operation simulation

capability is used to implement low-cost trial-
and-error of operations such as dynamic elastic
capacity expansion, realizing seamless service
operation changes.

3.2.3 IntelligentWAN — IP ADN implement full-lifecycle automatic scheduling of

network traffic, helping operators build leading
Huawei IP ADN solution has a set of key next-generation E2E intelligent IP networks.
components, including intelligent IP network
routers (such as NetEngine, NE, ATN, and CX
series) and intelligent network management and
control system (iMaster NCE). For IP networks,
iMaster NCE offers Path Computing Element,
intelligent O&M, and open programmability to

47
ADN

Cloud-network
one stop operation

Network Service APIs

Smart Real-time
Decision-making Awareness

IPv6+
Precise Control Public
Cloud

Private
Cloud

+SRv6 +IFIT +Slicing +AI

Industry
Cloud

Figure 3-5 IntelligentWAN

Highlights and self-healing IP transport network with the

best possible service experience.
1. Path Computing Element and One Connection
for Visualization reduce integration complexity 3. O p e n p ro g ra m m a b i l i t y a cce l e ra t e s
by 50%. Multidimensional indicators such as service rollout with "three 1s". Based on
latency and bandwidth utilization are displayed
YANG models, iMaster NCE provides end-to-
on the network digital map in real time. During
end open programmability, including device
path computation and optimization, up to 15+
driver programmability, network service
factors are flexibly combined to automatically
programmability, and open APIs for devices
schedule network traffic throughout the lifecycle.
and services. These capabilities make it possible

2. "Congestion Free" facilitates "Troubleshooting to complete prototyping in 1 week, testing in

0-1-3-5". "Troubleshooting 0-1-3-5" is a fault- 1 month, and commercialization in 1 quarter.

oriented closed-loop solution in which "0-1-3- More importantly, operators' O&M personnel
5" refers to the minutes needed for "prevention- are empowered to don the IT hat.
identification-demarcation/locating-recovery".
This solution aims to build a self-optimizing Key UCs

48
 ADN

1. Network digital map: Thanks to technologies network planning. In particular, by analyzing

like BGP-LS and Telemetry, indicators four typical types of congestions as well as
such as network latency, bandwidth, and the impact of transport network KPIs on base
utilization are visualized in a topology, in stations, iMaster NCE proactively identifies
real time. Navigation-like path computation bottlenecks in transport networks, which can
and one-stop provisioning in a map make be used as the basis for network planning.
configuration simple and efficient. Network
paths are automatically optimized based 5. Network slicing: Up to 10,000 slices can
on SLAs to continuously guarantee user be planned and automatically deployed
experience. with a single click. Network-wide slices
are visualized to facilitate management.
2. Network risk analysis: In the big data + If necessary, they can be losslessly scaled
three-layer AI architecture (device/network/ up or down with a single click. Automatic
cloud), devices are built with AI chips for management is woven throughout the entire
real-time awareness, networks leverage lifecycle of slices, providing differentiated SLA
AI models to construct local knowledge assurance in the 5G and cloud era.
bases, and the cloud provides an AI training
platform for federated learning. All this 6. Intelligent incident management: Huawei's
enables risk-centric predictive O&M that massive O&M data is trained online and
features configuration check, device KPI intelligent algorithms are used to transform
detection, device resource prediction, strong/ the focus of O&M from being based on a
weak signal analysis, co-routing impact massive number of alarms to being based on
analysis, and risky link identification. "one incident, one ticket". This increases alarm
compression rate to 99%+ and results in much
3. VPN service assurance: The IFIT technology higher ticket dispatch accuracy and O&M
visualizes tenant-level service SLAs and efficiency, as well as a better user experience.
provides differentiated private line services.
In addition, VPN service topology restoration 7. Agile Open Container (AOC): Users can
and KPI association analysis are used to define service YANG models themselves, and
demarcate faults in minutes, quickly identify programming is as easy as constructing toy
fault points, and accurately dispatch trouble blocks. Whenever a new service is needed,
tickets. it can be quickly developed and rolled out,
achieving the agile development goals of
4. Mobile transport congestion analysis: prototyping in 1 week, testing in 1 month,
From the perspective of base stations, iMaster and commercialization in 1 quarter. In
NCE displays the quality trend of network- addition, standard northbound APIs and
wide/regional base stations, and supports southbound device configuration scripts are
E2E topology restoration and SLA association automatically generated based on YANG
analysis in order to identify congestion points models, enabling minutes-level automatic
and provide root causes and optimization provisioning of end-to-end services.
suggestions. From the perspective of
transport networks, iMaster NCE proactively
identifies congestion bottlenecks and guides

49
ADN

3.2.4 IntelligentCampusNetwork — Enterprise Campus ADN

Huawei's ADN solution for campuses has three main components: CloudEngine campus switches,
AirEngine access points (APs), and iMaster NCE-Campus. Providing management, control, and analysis
functions for campus networks, the solution delivers full-lifecycle automated network management and
intelligent O&M services, leading the industry in implementing Level 3 ADN.

Application
layer
Authentication and Enterprise Commercial Enterprise
accounting O&M Wi-Fi office

Management Intent-driven Intelligent fault

layer recommendation remediation

Simulation and Information

Intelligent network
Automated network verification collection/analysis

Digital twins AI inference

Digitization

Network
layer

Simple-service Multi-service Multi-branch

campus campus interconnection campus

Terminal
layer POS Security Intelligent Smart Temperature
Printer terminal PC protection VR terminal meter sensor

Figure 3-6 IntelligentCampusNetwork

Highlights controlled. In contrast, Huawei's ADN supports

automatic terminal access, achieving second-
1. Z e ro - w a i t : N e t w o r k m a n a g e m e n t i s level seamless access.
automated, no longer relying on expert
experience. This ensures zero wait time for 3. Zero-interruption: In terms of network
network provisioning and reduces OPEX by O&M, passive responses are transformed to
85%. minute-level automatic optimizations without
service interruption.
2. Zero-intervention: On traditional networks,
terminal access needs to be manually Key UCs

50
 ADN

1. Intent-driven deployment : With the 5. 360-degree network health: Based on

scenario knowledge base and collaborative the evaluation of network health status
recommendation algorithm, iMaster NCE- from various dimensions, iMaster NCE-
Campus accurately maps service scenarios Campus can proactively identify more than
with networking models and recommends 200 typical problems and reduce potential
optimal network solutions by analyzing network problems by 85%.
service intents. In this way, networks can be
provisioned in minutes, without the need for 6. 360-degree user experience: iMaster NCE-
professional personnel. Campus can visualize the journey experience
of each user at any time, reducing user
2. Intelligent verification: iMaster NCE- complaints by 90%.
Campus provides a much faster network
change verification method than 7. 3 6 0 - d e g re e a p p l i c a t i o n a s s u ra n ce :
conventional options. Normally, a change With the capability to identify more than
takes just 10 minutes but verification lasts for 1000 mainstream applications, iMaster
4 hours, and there could be many complaints NCE-Campus can intelligently detect the
if verification is not sufficient. With iMaster application quality and reduce the time for
NCE-Campus, verification takes just minutes locating a fault from hours to minutes.
and assures changes without errors.
8. 360-degree intelligent optimization:
3. Free mobility: In Huawei's ADN solution In routine network optimization, iMaster
for campuses, free mobility is orchestrated NCE-Campus analyzes and predicts future
using natural language, achieving matrix- AP loads based on data of the previous
based simplified management, cross-vendor seven days and implements intelligent
deployment through IP-security group and automatic network optimization. It
mapping, and one-off policy configuration. also continuously trains terminal profiles
This delivers consistent service experience to and differentiates roaming steering for
users anytime and anywhere on all-wireless terminals based on AI algorithms, such as
campus networks. reinforcement learning. This improves the
roaming success rate by 70% and roaming
4. Intelligent terminal management: iMaster speed by 30%.
NCE-Campus supports intelligent terminal
management. It features a conventional
terminal fingerprint database and innovative
application AI clustering identification
capabilities. Such strengths allow iMaster
NCE- Campus to accurately identif y
networked terminals (98% of known types
and 95% of unknown types), allocate them
with correct networks, and prevent access
by spoofed terminals. This reduces manual
intervention and assures secure access in
seconds.

51
ADN

3.2.5 IntelligentFabric — Data Center ADN

Huawei's ADN solution for data centers consists of CloudEngine data center switches and
iMaster NCE — Huawei's autonomous driving network management and control system.
This sophisticated solution unifies the computing, storage, and service networks with an all-IP
architecture. It is also IPv6-capable and offers unique Level 3.5 ADN capabilities. What's more,
this future-proof solution supports full-lifecycle automation and network-wide intelligent O&M for
data center networks, reducing OPEX by 30% and enabling intelligent upgrades for enterprises.

Enterprise ITSM

Leaf node Policy Server Cross-cloud Network

capacity expansion change login interconnection path visibility

Service design platform

Intent-driven recommendation E2E visibility

Simulation & verification Fault remediation

Generic NE model

Unified data base

Public
cloud Vendor A

Public Public
cloud cloud Vendor B

SDN SDN Traditional Multi-cloud Multi-vendor

(network overlay) (host overlay) network network device

Figure 3-7 IntelligentFabric

Highlights helps prevent faults caused by manual errors

for maximized productivity.
1. Zero-wait: The service deployment and rollout
time is slashed to minutes, greatly reducing 3. Zero-interruption: Network trend prediction
the O&M workload and improving the O&M prevents network quality deterioration and
efficiency. failures, minimizing the possibility of faults.
Even though a fault occurs, it can be quickly
2. Zero-error: Evaluating network change risks and accurately demarcated and located.

52
 ADN

Key UCs and source tracing are performed, locating

root causes within 3 minutes. And thanks to
1. Intent-driven planning and deployment: the intelligent decision-making system, the
By understanding and translating customers' impact of faults is analyzed and the optimal
service and network intents, iMaster remediation solution is recommended,
NCE automatically selects the optimal rectifying typical faults within 5 minutes.
network deployment solution, enabling
E2E automated service provisioning and 4. Open industry ecosystem: Huawei's ADN

automated closed-loop management of is seamlessly integrated into enterprises'

intents throughout the lifecycle. O&M systems, becoming a key part of the
O&M process. This helps achieve automated

2. Network change simulation and evaluation: closed-loop management on the entire data

Based on the device configuration, topology, center network. In the northbound direction,

and resource information on the live network, iMaster NCE uses the runbook service
designer to flexibly orchestrate service flows
iMaster NCE uses network modeling and
and seamlessly interconnect with enterprises'
formal verification algorithms to evaluate
O&M systems. In the southbound direction,
network change risks and eliminate design
AOC — Huawei's open programmability
logic defects and configuration errors, thereby
platform — implements fast adaptation of
ensuring 100% configuration correctness.
devices from multiple vendors, achieving
automated provisioning of heterogeneous
3. "1-3-5" intelligent O&M: Telemetry technology
multi-vendor and multi-cloud networks in
is used to collect network performance data
minutes. On top of that, full open network
and comprehensively evaluate network health
data services facilitate integration with the
based on service experience to proactively
service performance monitoring system,
identify more than 60 potential risks and
implementing unified O&M of services and
over 90 typical faults within 1 minute.
networks.
Based on knowledge graphs and Huawei's
unique AI algorithms, fault aggregation

3.2.6 IntelligentFAN — All-Optical optical access networks (comprising FTTR,

Access ADN Digital QuickODN (DQ ODN), as well as

The core components of Huawei's all- smart ONT and OLT) and the iMaster NCE
optical access ADN solution include the all- intelligent management and control system.

53
ADN

Through digital modeling and introduction of application scenarios and high- qualit y
cloudification, big data, and AI technologies broadband experience. Against this backdrop,
for all-optical access networks, ODNs and single-domain autonomy of all-optical access
home networks are visualized and full- networks can be achieved. The following
lifecycle intelligent O&M is provided. This figure shows the solution architecture.
helps to meet requirements of differentiated

Intelligent
service Third-party Premium
Portal/APP
BSS/OSS broadband ADO

O&M/operations platform
Open APIs

Manager Controller Analyzer

Intelligent
Network Access

Intelligent management
and control system

Single-domain
autonomy

FTTR-Home

FTTR-SME
Intelligent
NE DQ ODN Smart OLT

All-optical access network

Figure 3-8 IntelligentFAN

Benefits network experience

1. Zero-wait service provisioning: DQ ODN, 3. Zero-touch service: Insights into STAs,

one-stop FTTR planning and acceptance, networking, applications, and experience
accurate resource information, and quick bottlenecks, user self-service of home
service provisioning network, precision marketing, and auxiliary
operations
2. Zero-trouble service: Intelligent O&M, self-
identification, self-diagnosis, and self- Key UCs
optimization of PON and Wi-Fi faults and
experience issues, and guaranteed broadband 1. Visualized ODN resource management:

54
 ADN

Based on image recognition and the DQ management and control system.

ODN optical iris solution, ODN topologies
can be automatically analyzed and 5. Precise identification of potential HBB
restored, ODN networks are visualized and subscribers: Insights into applications,
manageable, and the resource information is experience, networks, and bottlenecks help
accurate. identify potential subscriber requirements
and generate labels for potential subscribers.
2. Remote ODN acceptance and precise The insights also help operators identify
locating: Intelligent boards on OLTs and the b ro a d b a n d p a c k a g e u p g ra d e s , W i - F i
optical iris solution enable automatic and networking, and scenario-based potential
remote ODN network acceptance and optical subscriber requirements, greatly improving
line locating at the meter level on iMaster marketing success rate and improving the
NCE. quality of HBB.

3. Automatic identification of poor-QoE Home 6. One-stop FTTR acceptance: The FTTR field
Wi-Fi: Home Wi-Fi faults make up a large service app provided in the solution can be
proportion of home broadband faults. With used to perform one-stop acceptance on
intelligent boards on OLTs, the all-optical more than 10 broadband KPI types (such as
access ADN solution can capture undistorted networking, rate, delay, and roaming), and
traffic features. In addition, with the automatically generate acceptance reports to
unique intelligent Wi-Fi experience analysis ensure networking standardization as well as
algorithm, the solution comprehensively eliminate further home visits.
identifies faults including poor Wi-Fi coverage
and interference of network-wide subscribers, 7. Proactive FTTR assurance: T h e F T T R
visualizing Wi-Fi experience faults. "Optical+Wi-Fi" networking integrates
innovation in two aspects. First, remote
4. Broadband poor-QoE root cause locating: management of network-wide topologies
The first segment-based speed test capability and device statuses and the 24/7 playback
is used to accurately demarcate poor-QoE function are supported. These functions,
faults between STAs, Wi-Fi router, ONT, which are similar to CT scans, detect network
ODN, OLT, and bearer network. In addition, deterioration and facilitate troubleshooting.
the Huawei- developed spatiotemporal Second, the speed test assurance solution
association intelligent diagnosis algorithm is enables O&M personnel to remotely test the
used to identify more than 30 types of home network speed between STAs, slave FTTRs,
broadband (HBB) poor-QoE root causes the master FTTR, and Internet segments
of Wi-Fi network congestion, 100BASE-T with one click to identify potential risks
cables, and weak optical signals on optical of substandard bandwidth. In this way,
lines, greatly reducing workload for O&M network faults can be rectified before they
engineers, who would otherwise need to are perceived by subscribers, ensuring user
go door-to-door for diagnosis and fault experience.
check. Instead of carrying out home visits,
O&M engineers can remotely optimize and
rectify non-hardware faults on the intelligent

55
ADN

3.2.7 IntelligentOTN — All-Optical of all-optical networks and introduction of

Transport ADN cloudification, big data, and AI technologies,
dumb fiber resources are visualized, and an
The core components of Huawei's all-optical intelligent, green, and highly-reliable all-
transport ADN solution include the all-optical optical base is built, providing a premium
transport networks (such as OXC and OTN) private line service experience to customers.
and the iMaster NCE intelligent management
and control system. Through digital modeling

Service
intelligence BSS OSS Portal/Apps

O&M/Operations platform
Open APIs

Manager Controller Analyzer

Network
intelligence Transport

Intelligent management and control system

Single-domain
autonomy
2B

2C
C

NE 2H
intelligence

All-optical transport network

Figure 3-9 IntelligentOTN

Highlights display of optical cable networks, eliminating

the problems of managing dumb resources
1. Visualization of dumb resources: Visualizes through visualization.
and predicts fiber health status, locates
faults, intelligently identifies optical fibers 2. Intelligent and simplified O&M: Covers the
in the same cable, and provides GIS-based entire lifecycle of planning, construction,

56
 ADN

maintenance, optimization, and operations, and optimization suggestions are provided to

transforms O&M from passive to proactive, improve availability.
and builds networks with 99.999% reliability,
providing ultra-high-bandwidth, low-latency, 4. Intelligent identification of optical fibers in
and highly-reliable transport capacity to the same cable: Built-in intelligent sensors of
customers. optical NEs detect and analyze performance
data changes caused by the Rayleigh
3. Intelligent premium experience: Provides scattering effect, Brillouin scattering effect,
visualized and manageable SLA performance, and Raman scattering effect of optical fibers
a utomatic provisioning of private line in real time, and automatically identify risks
services, and after-sales SLA assurance. of working and protection fibers in the same
cable. In this way, services can be rectified to
Key UCs eliminate potential accidents.

1. Optical network resource assurance: Optical 5. Intelligent incident management: Three-

network resource data and planning and level intelligent alarm compression (alarm
design data are managed and analyzed compression, alarm aggregation, and
in a unified manner to implement unified root cause analysis) greatly improves
visualization, analysis and forecast, online troubleshooting efficiency, shortens the
check, and capacity expansion planning of duration of fault detection, diagnosis, and
network resources. This helps customers recovery, and enables "one ticket for one
learn about resource usage, identify network fault".
resource bottlenecks in advance, and
accurately expand capacity, realizing zero- 6. Latency map: A microseconds-level, real-
wait time for resources and shortening the time, and dynamic network-level latency map
service TTM. makes private line latency, which could not
be perceived or measured earlier, perceivable
2. Optical network health assurance: For OTS and accurately measurable. Similar to web
fiber and OCh performance deterioration mapping service applications, the latency
faults on WDM networks, intelligent forecast map allows operator marketing personnel
and analysis of optical network subhealth to evaluate whether the latency and
status are performed based on seconds-level bandwidth between sites meet tenants'
awareness and edge inference capabilities requirements. This facilitates mapping
on the device side, achieving visualized and of network resources and marketing of
predictable fiber and OCh health status private line services that have differentiated
and fault locating, transforming O&M from SLAs.
passive to proactive, and reducing potential
risks of optical line interruptions. 7. Agile service provisioning: Support is
available fo r multiple service scenarios
3. Availability assurance: Availability of fibers (including Client, EoO, EoS, SDH, and
and services are automatically evaluated and MPLS-TP services) and automatic service
analyzed, potential risks and bottlenecks to configuration with only specified source
availability on the live network are identified, and sink ends. Network capabilities are fully

57
ADN

opened through standard ACTN APIs, and private line services is monitored in real
integration with the OSSs/BSSs is simplified time. Warnings are generated when the
for quickly integrating NCE into carriers' performance indicators exceed thresholds,
private line service production processes, so that potential risks can be eliminated and
implementing automatic provisioning of there is no breach of contract.
private line services, and improving the self-
service level.

8. Private line SLA: The SLA performance (such

as latency, bandwidth, and availability) of

3.2.8 IntelligentServiceEngine — AUTIN: Intelligent O&M

Intelligent O&M
Commercial use of 5G networks, large-
Huawei IntelligentServiceEngine implements scale complex networks, and diversified new
closed-loop management to meet operators' services pose many O&M challenges for
service objectives and domain objectives, and operators in terms of cost, efficiency, quality,
facilitates systematic, scenario-specific digital and talent transformation. Operators' digital
transformation for operators. It consists of intelligent transformation for network O&M is
various key modules, including the domain speeding up due to the COVID-19 pandemic
knowledge engine, business intelligence and has become an industry trend.
engine, hyper-automation engine, domain
app development engine, and network and AUTIN, an intelligent Huawei O&M solution,
environment digital twins. Based on the digital meets three key O&M requirements: quality
intelligent transformation practices from its improvement, efficiency improvement, and
partnership with over 100 operators around rapid talent upskilling. The solution helps
the world, Huawei has launched three digital operators quickly implement digital intelligent
intelligent O&M solutions (AUTIN, SmartCare, transformation towards Zero-X network O&M.
and ADO) to achieve efficient O&M, deliver
a superior experience, and facilitate agile Quality improvement: In 5GC scenarios,
businesses. major faults have a large impact scope, and

58
 ADN

IntelligentServiceEngine

User Intelligent Scenario- Virtual

experience AD business enabler specific O&M AD employee

Business Hyper-
intelligence automation
Domain app
Domain engine engine
development
knowledge
Cross-domain (process) E2E cross-domain engine
engine collaboration spanning (network)
time and space collaboration
Security
management

Network and environment

Network enabler
digital twins

Figure 3-10 IntelligentServiceEngine

it takes a long time to demarcate and locate O&M talent upskilling: O&M personnel need
them. Huawei AUTIN intelligently predicts to upgrade their skills after O&M operations
service impacts and risks, and automatically are automated. Huawei's open platform and
diagnoses faults within just a few minutes, abundant O&M knowledge assets enable
significantly cutting the MTTR. orchestratable, low-code O&M application
development, simplifying O&M talent
Efficiency improvement: Over 70% of daily upskilling. The systematic Accompanying

operations in a network operations center Service and enablement service for talent
upskilling are provided to shorten the O&M
(NOC) are repetitive, inefficient, and depend
application development period from months
heavily on expertise. Huawei AUTIN helps
to weeks.
NOCs implement integrated, automated
monitoring and maintenance, intelligent
AUTIN continuously integrates expertise
diagnosis, and self-service, closed-loop
and iteratively optimizes domain-specific
management of onsite operations through
automated, digital, intelligent engines
mobile phones. Faults can be automatically to improve its level of automation and
diagnosed and managed in a closed-loop intelligence.
manner, reducing the number of tickets per
NE and greatly improving O&M efficiency. SmartCare: Superior Experience

59
ADN

Delivering a scenario-specific, superior user in order to maximize the usage of network

experience is the backbone of development resources, help operators achieve a high 5G
strategies for most operators. This goal helps network experience ranking, and improve their
operators deploy high-performance networks, market share.
achieve business success, and maintain a
leading market position. The convergent data platform of Huawei
SmartCare converges and analyzes data from
Scenario 1: Optimal voice over LTE (VoLTE) multiple domains. Built-in intelligent operators,
experience. Operators need to remove out-of-the-box intelligent prediction models, and
2G and 3G networks, and deploy a high- abundant expertise models help operators create
quality fundamental VoLTE network. Huawei digital twins to improve network experience,
SmartCare performs multi-domain, multi- and achieves efficient data-driven network
interface data association analysis and operations, superior network performance, and
slicing to accurately locate VoLTE experience leading business results.
problems and provide proactive assurance,
preventing surges in user complaints. The ADO: Business Enablement
VoLTE experience assurance solution helps
operators greatly reduce complaints regarding ADO is a premium home broadband solution.
voice services. The convergent data platform Based on in-depth awareness of user-level
associates wireless network, core network, service experience indicators, ADO models
terminal, and registration information to services for poor-QoE users, causes of poor QoE,
facilitate VoLTE user migration. and potential users. It provides functions such
as Internet access quality analysis, single-user
Scenario 2: Optimal 5G experience. Quickly quality query, potential user identification, and
increasing 5G users and traffic is the key VIP experience assurance to support closed-loop
for operators looking to create a positive poor-QoE rectification and precision marketing.
5G business cycle. Algorithms are improved
to focus on optimal experience instead of Ultimate home broadband experience:

coverage in order to allow more users to camp This solution helps operators build a home
on 5G networks and maintain the optimal rate broadband user experience model based on the
for these users. SmartCare helps operators analysis of user complaints, user experience,
significantly improve the camping ratio. The user behavior, and Internet access quality, and
convergent data platform and the insights about allows operators to add service tags to models.
mobile phones, networks, and packages help The prediction accuracy for poor QoE is 80%,
operators attract more target 5G users and and operators can implement proactive O&M to
increase the proportion of 5G traffic. eliminate potential faults, reduce the complaint
rate, and improve user experience.
Scenario 3: Highest experience ranking. Network
experience ranking is the key to the brand image G r i d - b a s e d p re c i s e o p e ra t i o n s : To m e e t
of operators. Traditional optimization based on differentiated service requirements of individual
drive tests is inefficient. SmartCare makes use of users, family users, governments, and enterprises,
third-party crowdtest data to perform analysis this solution analyzes and predicts the network
and device-pipe-cloud collaborative optimization bearer capability based on grid-based service

60
 ADN

insights, matches the resource capability of the and rectification help realize precise grid-based
live network, and allows operators to expand service development.
capacity and plan the network in advance. In
addition, grid-based poor-QoE identification

3.2.9 ADN Evolution Path and Method

AN evolution is a long-term process. The first AN white paper released by TM Forum in 2019 provided
a phase-by-phase evolution from manual O&M Level 0 to full autonomous Level 5.

L0: Manual L1: Assisted L2: Partial L3: Conditional L4: High L5: Full
Level
operation& operation& autonomous autonomous autonomous autonomous
Definition
maintenance maintenance network network network network

Execution p P/S S S S S

Awareness p P/S P/S S S S

Analysis p P P/S P/S S S

Decision p P P P/S S S

Intent/
p P P P P/S S
Experience

Applicability N/A Select scenarios All scenarios

P: Personnel, S: Systems

Table 3-1 AN levels

61
ADN

Level 0 - Manual O&M: The system delivers Level 4 - High autonomous: Based on Level
assisted monitoring capabilities, which means all 3 capabilities, the system performs analysis
dynamic tasks have to be executed manually. and makes decisions based on predictive or
active closed-loop management of services and
Level 1 - Assisted O&M: The system executes customer experience-driven networks in more
certain repetitive sub-tasks based on pre- complex cross-domain environments.
configurations to improve execution efficiency.
Level 5 - Full autonomous: This level is the
Level 2 - Partial autonomous: The system ultimate goal for telecom network evolution.
enables closed-loop O&M for certain units based The system provides closed-loop automation
on pre-defined rules/policies under certain capabilities across multiple services, domains,

external environments. and the entire lifecycle.

Level 3 - Conditional autonomous: Based on To h i g h l i g h t t h e c o m m e rc i a l v a l u e a n d

Level 2 capabilities, the system senses real- measure customer experience, Huawei ADN
solution further defines the vision, objectives,
time environmental changes, and optimizes
and roadmap for each level based on the TM
and adjusts itself to the external environment in
Forum's ANL definitions.
certain network domains.

Uninterrupted Quality Experience beyond

Zero-trouble
services assurance expectation

Targeted Intelligent
Zero-touch visits Self-service recommendation

Acceptable Predictable Committable

Zero-wait duration duration duration

Level 2 Level 3 Level 4

ANL 2021 2023 2025

Device Network Uninterrupted services

Self-configuration self-management self-provisioning during upgrade

Fault Fault Fault

Self-healing self-identification self-diagnosis self-healing

Scenario-specific Collaborative
Self-optimizing Offline optimization
optimization optimization

Providing Zero-Wait, Zero-Touch, and Zero- Self-service level: Reach Level 3 by 2023 to
Trouble Innovative Network Services for achieve full self-service, and reach Level 4 by
Individual and Industry Customers 2025 to provide intelligent recommendation
services.
Service provisioning and change duration: Reach
Level 3 by 2023 to reduce the duration to an Customer experience assurance: Reach Level 3 by
acceptable level, and reach Level 4 by 2025 to 2023 to provide quality assurance, and reach Level
guarantee a committed duration. 4 by 2025 to deliver the best customer experience.

62
 ADN

Deploying Self-configuration, Self-healing, Network performance and quality optimization:

and Self-optimizing High-Quality Network Reach Level 3 by 2023 to perform scenario-
O&M and Infrastructure specific optimization, and reach Level 4 by 2025
to use various methods to perform collaborative
Network deployment and service provisioning: optimization for multiple objectives.
Reach Level 3 by 2023 to implement self-
provisioning for basic network services, and Four Elements Key to Solid Industry Development
reach Level 4 by 2025 to ensure service
continuity during network device upgrades. Based on the overall evolution path and
objectives, Huawei has summarized four key
Fault diagnosis and rectification: Reach Level 3 elements after joint research on AN with leading
by 2023 to implement self-detection and self- operators over the past two years.
diagnosis of most faults, and reach Level 4 by
2025 to achieve self-healing for common faults.

Vision and Vision Target architecture

objectives

ANL Effectiveness indicators

Generic Domain- Closed-loop Zero-X

ANL specific ANL evolution
Evaluation method Self-X

Operations Overall implementation ANL evaluation and Application innovation and Capability development
practices planning weakness identification pilot verification and application promotion

Figure 3-11 AN evolution

1. Target architecture: Based on TM Forum's comprehensively and objectively measure

target architecture of AN and operators' the ANL, identify weaknesses, formulate
business strategies and objectives, a mid- improvement strategies, launch technological
and long-term target architecture is defined research, and develop innovative applications.
to facilitate AN development from top to
bottom. 3. Effectiveness indicators: The business value
and service effectiveness are defined for each
2. Level standards: Under the target architecture, ANL to explicitly and quantitatively measure
generic ANL, domain-specific ANL, and AN evolution, ensuring AN capabilit y
evaluation methods are developed to development meets enterprise strategies and

63
ADN

service development goals. Both effectiveness evaluation, weakness identification, pilot

indicators and ANL are required to fulfill the verification, capability development, and
AN vision. application promotion, facilitating the
evolution towards Level 5.
4. Operations practice: Operations practices
are iteratively carried out and summarized
through overall planning, level-based

64
04
Outlook and
Initiatives

65
 ADN

IP technology reconstructed the forwarding work together to evolve towards high-level AN. In
architecture of the communications network this regard, we propose the following suggestions:
two decades ago. Cloud technology profoundly
transformed the network management and control 1. Focus on innovation, revenue growth,
architecture a decade ago. In the next decade, AI quality improvement, and efficiency
technology will be introduced into each layer in improvement, and continuously develop
the network architecture to promote evolution innovative applications: We recommend that
toward intelligent networks. NE status data and AI all industry partners focus on innovation,
need to be integrated, and innovative algorithms revenue growth, quality enhancement, and
need to be developed to approach the theoretical efficiency improvement, and continuously
limit, transforming uncertainty to certainty and develop innovative applications for planning,
improving network performance. Network O&M construction, maintenance, optimization,
data and AI also need to be integrated, and big and marketing in all network domains,
data analysis and closed-loop optimization are a n d i n h i g h -va l u e s c e n a r i o s , s u c h a s
required to comprehensively improve the level 5GtoB, enterprise cloud access, and home
of network automation and service capabilities broadband, contributing a host of mature
in all scenarios. Edge intelligence is required to commercial solutions for AN.
detect diversified service requirements in various
industries and improve service experience. 2. Encourage operators to accelerate large-
scale Level 3 deployment to create a positive
Communications networks are full of potential business cycle and advance the industry:
as well as uncertainties. Standards and industry According to the latest survey of TM Forum,
organizations, operators, suppliers, research institutes, 37.1% of the respondents believe that AN
and industry management departments need to will reach Level 3 over the next three years

66
ADN

and create significant business value. We innovation with upstream and downstream
recommend that operators speed up the industry partners.
process of introducing and experimenting new
AN technologies, products, and solutions, and 4. Jointly define Level 4 target profiles, launch
the process of replication and promotion of basic theoretical research, and make
mature technologies, products, and services. We technological breakthroughs: AN evolution
also suggest that they accelerate the process from Level 3 to Level 4 is a huge progress
of using autonomous capabilities to deliver that creates tremendous value but is also
business value and operations benefits, seize fraught with challenges. All parties in the
new opportunities for digital transformation, industry should focus on business and
and promote a virtuous business cycle for service requirements, target architecture,
industry development. key technologies, core capabilities, evolution
path, and product and solution innovation,
3. Establish an industry collaboration platform and jointly define Level 4 target profiles to
to co-develop and approve standards, and continuously improve AN capabilities, such
promote efficient collaboration among the as accurate real-time detection in complex
industry, academia, research institutes, and scenarios, real-time simulation based on
business circle: We recommend that M-SDOs digital twins, and data-driven user experience
further collaborate with each other and utilize analysis and decision-making.
a top-level design and efficient communication
mode to achieve standards recognition The large-scale deployment of AN calls for close
between different network domains and collaboration across the industry. Huawei is willing
establish a standard co-construction and to work with industry partners to launch more
mutual recognition mechanism. They should basic theoretical and technological research, and
also accelerate the transformation from leverage Huawei's technological advantages in
industry standards to business and technical the ICT field and further contribute to standards,
implementation solutions, achieve efficient technologies, and business innovation, promoting
collaboration between industry, academia, high-quality network development toward building
research, and application, and promote joint an intelligent world.

67
HUAWEI TECHNOLOGIES CO., LTD.
Huawei Industrial Base
Bantian Longgang
Shenzhen 518129, P.R. China
Tel: +86-755-28780808
www.huawei.com

Trademark Notice
is a trademark or registered trademark of Huawei Technologies Co., Ltd.
Other trademarks, product, service and company names mentioned are the property of their respective owners.

General Disclaimer
The information in this document may contain predictive statements including, without limitation, statements
regarding the future financial and operating results, future product portfolio, new technology, etc. There are a
number of factors that could cause actual results and developments to differ materially from those expressed
or implied in the predictive statements. Therefore, such information is provided for reference purpose only and
constitutes neither an offer nor an acceptance. Huawei may change the information at any time without notice.

Copyright © Huawei Technologies Co., Ltd. 2022. All rights reserved.

No part of this document may be reproduced or transmitted in any form or by any means without
prior written consent of Huawei Technologies Co., Ltd.