HUAWEI Intelligent Operations Solution/media/CNBGV2/download/products/serv… · HUAWEI's...

2Trends and Challenges in Network Operations

HUAWEI Intelligent Operations Solution White Paper

Market trends for network operations 011.1

Challenges in network operations 021.2

The urgency of operations and maintenance transformation 071.3

Deploying AI is the trend in industry 111.4

Trends and Challenges in Network Operations1

Requirements for the future operating model 132.1

The role of AI in telecom network operations and maintenance 142.2

Future model of network operations: "Human + Machine" Collaboration Model 152.3

Business benefits of the "Human + Machine" Collaboration Model 172.4

The evolution path to the "Human + Machine" Collaboration Model 182.5

Future Operating Model for Network Operations2

Design principles of HUAWEI's intelligent operations and maintenance service 193.1

Solution architecture 233.2

Core capabilities 253.3

Transformation plan to HUAWEI's intelligent operations platform 293.4

Intelligent Operations of HUAWEI3

Large operator in China: Automatic fault management 314.1

Intelligent operations & maintenance transformation of a large Southeast Asia telco 334.2

Success Stories4

Summary and Outlook5

References6

CON

TENTS

1 Trends and Challenges in Network Operations

The market and technology have driven the evolution of telecommunication industry, bringing disruptive changes and new requirements to communication networks.

1.1 Market trends for network operations

Trends and Challengesin Network Operations1

new business products and services based on new technology are continually launched (IoT, HD video, video calls, VoLTE, vertical business services, cloud services, etc.), requiring new operations systems and processes. The implementation time for these new products and services needs to meet the demands of their expected lifecycle. As one of the key business enablers, network operations must have better flexibility and efficiency so they adapt to the requirements of future business changes while being more robust in ensuring network quality and performance.

a more stable network environment along with more customized, differentiated, and high-quality operations services will be key factors for improving customer experience and customer loyalty.

in the past three years, the number of global physical stations has increased from 4 million to 6 million; meanwhile, the number of logical stations has expanded from 7.8 million to 15 million2, and the frequency spectrum has increased from three main frequency bands to more than ten, resulting in large-scale simultaneous usage of heterogeneous networks (LTE, UMTS, WLAN).

in the 5G era, operators are investing heavily in building infrastructure. Accompanied by the growth of CAPEX, OPEX increased dramatically (more than 20% in five years), while the operators' current revenue growth rate is less than 1% worldwide. Considering that OPEX accounts for 74%1 of operators' revenue, current revenue growth has been unable to support the rising cost pressure. In the future, reducing operations cost will become one of the key drivers for the transformation.

More diversified business

more flexible and efficient operations

More stable network

tighter focus on customer experience

More complex network environment

more intelligent operations and maintenance

Greater cost pressure

more cost-effective operations and maintenance


The development trend of telecommunication networks determines the need to design and deploy new network operating models supporting future business expectations, while current network operations are still too reactive, slow, and not agile enough to face the new expectations listed above. There is a significant gap between the current, outdated operating model and the expectations raised by the digital business and demanding end users.

The below figure shows how the major trends in the telecom industry affect the demands on network operations and how the new demand confronts the current situation, resulting in the following challenges.

Complex network environments bring a higher frequency of faults and significantly increase their complexity and resolution difficulty. Network maintenance requires not only efficiency and optimized cost of operations, but also new and more intelligent (i.e. AI-driven) operating models to support the complex and heterogeneous telecommunication services of the future.

Global business trends and revenue growth curves of operators

Fixed-line voice communication

60sFixed-linetelephone

70sFixed-line telephone

80sFixed-line telephoneData Communication

90sMobile PhoneInternet

2000-Mobile Internet

2010-Mobile Internet

2020-Internet of everything

Mobile voice communication + Short message

Mobile data+ Fixed broadband

Value-added services and OTT

Digital solutions

Platform Enabling

Infrastructure

Telecom service revenue

1.2 Challenges in network operations

Figure 1: Global business trends of operators


More diversified services

The development trend of telecommunication network

Current operations and maintenance situation

Requirements for operations and maintenance

The challenges operations and maintenance face

Greater cost pressure More attention paid to user experience

More complex network environment

Long operations and maintenance cycle, resulting in long online cycle of new business and high operations and high network operations and maintenance cost.

The current operations and maintenance mode, which relies heavily on manpower, will lead to further uncontrollable costs.

Poor user experience leads to faster loss of customers and revenue

It is difficult to solve problems effectively, and the maintenance period is longer.

More flexible and efficient operations and maintenance

More cost-effective operations and maintenance

Higher-quality operations and maintenance

More intelligent operations and maintenance

Separation of operations and

maintenance systems

Too many people involved

Passive operations and maintenance

Difficult fault resolution

Challenges brought by diversified business

Advanced network technology enables a plentiful spectrum of services covering all aspects of human life. At the same time, with the rapid development of innovative business, the industry's requirements for the launch of new products and services have changed from months to days.

Below, we show an overview of current business development trends.

Figure 2: The development trend of telecommunication network brings challenges to network operations and maintenance


End User

Consumer

FamilyEnterprise

Extension ofvalue chain

Traditional network service

Network technology enabling

Traditional network service

Extension ofvalue chain

End User

MassiveMIMO

RAN

SD WAN

NFV & vCPEVOS

FO BO

Traditional OSS

Personnel

Platform layer

Network

FME

Electronic operations and maintenance

Network optimization

Enterprise customers

5G

3G/4G

Home broadband

Fault management

Resourcemanagement

Performance management

Current operations and maintenance status: more than 60%3 of operators in the world are maintaining multiple isolated OSS systems. Business data and systems supporting mobile networks, fixed networks, and broadband are not fully integrated yet, resulting in poor resource efficiency, incomplete end-to-end operations capabilities, slow processes, and a high cost for new requirements customization. Siloed applications restrain agile and flexible operations. At present, 90%3 of the working time of operations personnel is spent on fault location, and that leads to low operational efficiency.The architecture of traditional network operations and maintenance systems is shown below.

Challenges that network operations and maintenance may face: in the context of business diversification, the lack of necessary flexibility in complex interfaces will lead to a longer operations service lifecycle, resulting in a longer timeline for the launch of new products and services, higher operational costs, and difficulty in effectively supporting business development.

CloudFoundation

SafetyCloud

BasicTelecommunication

Voice


DataCenter

EnterpriseNetwork

HomeEquipment

RouterGateway


OpticalfiberBroadband

ContentServices

Music

LiveVideo

Game

TerminalEquipment

POSMachine

MobilePaymentMobile

Phone

WearableDevices

Industry applications

IntelligentCity

MobileHealthIndustrial

Internet

Internet ofvehicles

Intelligent Home

SmartTV

SmartKitchen

IntelligentSpeaker

Message

Data

Figure 3: Panorama of current business trend

Figure 4: Architecture of traditional network operations and maintenance mode


Challenges brought by cost pressure

Challenges with customer experience management

In the past decade, the OPEX share in operators' revenue has increased from 64% to 74%4, with an average annual growth rate of 1%. Global operators have entered a period of slow revenue growth, with annual CAGR less than 0.5%, and negative profit growth. It can be predicted that with the further expansion of the number of global stations and frequency bands, operators will face greater OPEX pressure.

At present, customer behavior has significantly changed. Customers' loyalty to operators has lowered, and their overall satisfaction in the telecom industry is only 11% on average, at the bottom of the ranking in the NPS (net promoter score) survey5 of 15 industries and half that of the insurance industry. This will have negative effects on maintaining long-term stable customer relationships and on creating value. Today, the competition for telecom operators crosses industry boundaries. 54% of telecom executives believe that their future strongest competitors will come from other industries.

Under the pressure of increasing market competition and higher customer-acquisition cost, operators need to seriously focus on customer experience and provide better services.

20131.500

1.600

1.700

1.800

1.900

2.000

2.100

2015 20172014 2016 2018 2019

0,52%

2013 2014 2015 2016 2017 2018 2019

+23%

Source：GSMA intelligence, S&P,IHS Market Source：GSMA intelligence, S&P,IHS Market

Figure 5：OPEX growth of global operators 13-19 (unit: US $1 billion) Figure 6: Global operator revenue growth 13-19 (Unit: US $1 billion)

Current state of operations and maintenance: many operational activities of existing networks still heavily rely on human interventions. Daily operations and maintenance costs account for more than 70% of the total OPEX cost. The degree of automation is generally low. For example, the integration between different procedures are handled manually.

Because of more complex network environments and the exponential growth of network elements and devices in the future, there may be several network slices. Network slices are completely separated and independent layers designed to not affect other slices if something goes wrong in one of them. This separation and independence also enables adding new slices without impacting the rest of the network. Each network slice will have a specific topology structure and will need to manage a huge amount of equipment information.

The current operating model, which highly relies on human interventions, will generate more manual operations in the future, and further increase the operational costs.


customer experience management is still mainly reactive. 75%6 of network problems are discovered and raised by customers. 60%7 of telco carriers' operations lack full automation capacity and closed-loop operations processes for rapid problem solving. Faults can be very difficult to detect and prevent in advance, and customer experience and satisfaction are difficult to guarantee.

Current state of operations

37%9 of the failures are caused by network changes that increase complexity, and the current operating models heavily relying on human intervention and individual skills struggle to effectively handle the volume of problems. The demands of network management has exceeded people's knowledge and ability.

Current operations and maintenance status

with the continuous increase of network scale and data volume, the frequency of failures will be higher. The current operations struggle to meet users' demands for network quality and performance, and the current fierce market competition leads to potential loss of customers.

Challenges network operations may face

in the future, the number of interfaces will increase by 120%, and the number of nodes will grow more than 20 times.10 Fault frequency will also increase exponentially. With this perspective, the current traditional operating models relying on human experience and knowledge will become inadequate to manage incidents, meet the demand, and ensure the expected customer experience.

The challenges that network operations and maintenance may face

Challenges brought by more complex network environment

In the past three years, the number of physical stations has increased from 4 million to 6 million globally, meanwhile the number of logical stations has expanded from 7.8 million to 15 million, and the frequency spectrum has increased from 3 main frequency bands to more than 10.8 Under the existing EPC (Evolved Packet Core) network, and with the continual network development, the complexity of the network elements and their functions has gradually increased together with the number of the interfaces: from 2G and 3G to the 4G network, whenever a new network element is added, the interface of the network element and its compatibility with the existing network need to be managed.

The introduction of new network elements and interfaces may affect many other existing network elements, with further impacts on the end-to-end network maintenance costs.

Factors influencing broadband network subscription

Influencing factors

Overall experienceAdvertisementPublic praiseNetwork qualityDirect marketing

313022278

Consideration511813308

Evaluation6959278

Purchase

Figure 7: Factors influencing broadband network subscription

16% When choosing wireless network services, they have specific requirements for operators' brands

Willing to try new network services provided by digital subversives45%

Think the new network service will be better than the old one49%

Priority will be given to service providers that can provide fixed network, wireless and other forms of switching70%

Figure 8: Behavior statistics of user changing operators


Service demand = f1 (network complexity), where f1 fits the exponential function

• Network complexity = {number of connections, number of network elements, number of paths, number of B2B parties}

• Service demand = {number of faults, number of complaints, number of service requests, number of network changes}

O & M manageability = f2 (service demand) = f2 (f1 (network complexity))

• Network complexity = {number of connections, number of network elements, number of paths, number of B2B parties}

• Service demand = {number of faults, number of complaints, number of service requests, number of network changes}

Future service demand will explode after the turning point

Current position of telcos

Service Demand• Number of faults• Number of complaints• Number of service

requests• Number of network

changes

Network Complexity• Number of connections• Number of network nodes• Number of network routes• Number of B2B users

Service Demand• Number of faults• Number of complaints• Number of service requests• Number of network changes

Due to the diversification of network services and the plethora of application scenarios in the era of 5G, telecom operators are accelerating the deployment of network infrastructure to meet the demands of business development. Therefore, the number of connections, network elements, paths, and business partners will obviously increase. With the increasing network-environment complexity, the number of faults, complaints, service requests, network changes, and demands managed by operators will increase exponentially. The relationship between the two parts can be expressed by formula f1:

perations and maintenance manageability = {number of operations and maintenance staff, closed-loop processing time, number of human errors}

Consequently，the number of operational staff, overall time spent on addressing problems and new demands, and the quantity and probability of human errors will be amplified. The relationship between these two parts can be reflected by formula f2.

1.3 The urgency of operations and maintenance transformation

Manageability of Operations• Size of ops team• Closed-loop processing

time• Number of man-made

errors

Operations will be become unmanageable


Figure 9: F1 schematic diagram under passive operations and maintenance mode

Figure 10: F2 schematic diagram under passive operations and maintenance mode


Future service demand will explode after the turning point


Service Demand• Number of faults• Number of complaints• Number of service

requests• Number of network

changes

Network Complexity• Number of connections• Number of network nodes• Number of network routes• Number of B2B users

Service Demand• Number of faults• Number of complaints• Number of service requests• Number of network changes

Level of operations and maintenance

Capacities of operations and maintenance

Customer

Proactive management Predictive / defensive managementPassive management

Netw

orkService/Business

Customer-oriented operations and maintenance

Business- and service-oriented operations and maintenance

Network- and equipment-oriented operations and maintenance

• Take the initiative to monitor and measure customer perception; identify and improve weaknesses in combination with customer complaints

• Customer self-service• Highly automated service

• Service end-to-end management, automatic process management

• Support the arrangement and operations of digital business and services

• Support rapid iterative development of services

• Support the SRE operations and maintenance of applications and services

• Realizing the active discovery of performance and other faults through tools

• Suitable for traditional network and ICT infrastructure

• Based on data analysis and AI, proactively identify potential faults and prevent business interruption

• Applicable to traditional and virtualized networks

• Passive response to customer complaints

• Passive response after business interruption

• Traditional network, passive response

According to industry forecasts, the development trend of operations will gradually change from reactive and proactive management of network infrastructure to predictive and preventive management of customer experience. Self-service modes for customers will be provided through highly automated services.

Although network environments tend to be increasingly complex and diverse, through the customer-oriented and predictive/defensive intelligent operating model driven by AI, the processing time of network faults and the cost of operations, as well as human errors, can be effectively controlled.

Manageability of Operations• Size of ops team• Closed-loop processing

time• Number of man-made

errors

... But active management can easily meet the exploding demand

ReactiveManagement

Active Management

Figure 12: F1 schematic diagram under active operations and maintenance mode

Figure 13: F2 schematic diagram under active operations and maintenance mode

Figure 11: Transformation trend of network operations and maintenance


Includes 10-15%opex fromoutsourcing

50%

45%

40%

35%

30%

25%

20%

15%

10%

5%

0%

Tech

Hea

lthca

re

Telc

os

Cap

Goo

ds

Cons

truc

tion

Auto

s

HPC

/Lux

/Spo

rt

Food

Bev

s

Ener

gy

Reso

urce

s

Reta

il

Util

ities

Med

ia

Trav

el

The deployment of AI in operations within the telecom industry is still at the initial stage of development.

Since the concept of industry 4.0 was introduced in 201111, industrial giants in the manufacturing industry, such as Siemens and GE, have been gradually using AI to predict risks in the production process, reduce material waste and energy consumption, and improve overall production efficiency.

In the medical industry, intelligent health care has been able to collect data including step counts, blood glucose levels, heart rate, lifestyle and habits, and sleep quality through aggregation and processing to achieve monitoring, diagnosis, treatment, and other purposes.

In the media industry, AI programs can automatically write news, therefore not only the improving the timeliness of news but also reducing journalists' workload.

In the transportation field, the use of AI technology can analyze urban traffic flow in real time, adjust the interval between traffic lights, shorten the waiting time of vehicles, and improve the traffic efficiency of urban roads.

According to Goldman Sachs12 research (shown in the figure below), operational costs of the telecom industry are higher than traditional industries like automobile and construction, and most of the manual activities could be easily automated by exploiting AI's long-term potential.

Telecom operators are keen to design and adopt new operating models in order to meet the operational needs of the current physical networks and the future hybrid and virtual networks. Indeed, nearly 90%13 of the current operators have started or are getting ready to start operations-transformation projects. According to Gartner, intelligent operations and maintenance platforms will become the mainstream in the next five years. By 2022, 40%14 of large- and medium-sized enterprises will deploy intelligent operations platforms, significantly reducing operational costs while ensuring more stable, efficient, and safe operations of their business systems. At the same time, many leading operators are taking effective measures in multiple directions to carry out operations transformation in order to meet the current business development demands.

Source：Digital Divergence, How European Telcos can cut costs with automation, AI and big data, Goldman Sachs

Figure 14: Proportion of labor cost in various industries


Reduce OPEX: in the past five years, OPEX expenditure of global operators has increased by 20%15.

Many of the world's large telecom operators will increase investment in intelligent operations in the future to effectively improve operations efficiency.

Orange announced in their investors' conference at the end of 201716 that, in the future, it will promote the network automation strategy and save network OPEX with the help of big data and AI capabilities in order to reduce repetitive work, energy cost, rental cost, and idle time.

Improve user experience: currently, many telecom organizations have begun to invest in AI, big data analysis, robotic process automation (RPA), and other technologies to achieve fault "self-healing" and active maintenance.

Vodafone applies AI to network operations to predict network problems and preprocess them in order to reduce fault risk.

AT&T developed and launched the UNI (UAV Network Inspection) system. UAVs were introduced into the operations support systems to realize intelligent inspection of the towers so that they can conduct real-time monitoring and reduce faults.

China Unicom uses automation as well as other technologies to improve the execution speed of network operations and adopts the full set of automatic collection-analysis-execution, resulting in a 19% improvement of operations efficiency in the analysis process.

China Mobile uses PON (Passive Optical Network), for fault location and prediction as well as remote fault demarcation, location, and root cause analysis, so that the frequency of on-site visits is greatly reduced while the early warning and prediction accuracy of optical device and path degradation is significantly improved.

Promote the transformation of operations and maintenance: a new wave of transformation of the telecommunication industry is on the rise, and many operators have launched digital and intelligent transformation programs in the field of network operations, reengineering traditional organizations, processes, and platforms.

China Telecom has carried out an "intelligent development & transformation upgrading" plan to deploy SDN + NFV + Cloud to redesign the network operating model.

SK Telecom is deploying the TANGO platform, which runs as an independent OSS platform with a central intelligent platform, performing data collection, real-time analysis, decision-making, control functions, and intelligent troubleshooting.

At present, the leading enterprises in the market are transforming their network operations in many dimensions. Operators should embrace the current market trends, accelerate their transformation, increase operations efficiency, improve customer experience, and realize their business development plans.


Plateau will be reached:

As of August 2019

InnovationTrigger

Biotech – Culturedor Artificial Tissue

AR CloudDecentralized Web

Generative Adversarial NetworksDecentralized Web

Generative AdversarialNetworks

Adaptive ML-

Trough ofDisillusionment

Slope ofEnlightenment

Plateau ofProductivity

Expe

ctat

ions

Gartner Hype Cycle for Emerging Technologies, 2019

Time

Peak ofInflated

Expectations

Immersive Workspaces

Autonomous Driving Level 4

3D Sensing Cameras

Next-Generation Memory

Graph Analytics

5G

Biochips

Al PaaS

Edge AnalyticsAutonomous Driving Level 5

Low-Earth-Orbit Satellite SystemsExplainable AlPersonification

Knowledge GraphsLight Cargo Delivery Drones

Transfer Learning

Nanoscale 3DPrinting

DigitalOps

Flying Autonomous VehiclesAugmented Intelligence

Decentralized Autonomous Organization

Synthetic DataEmotion Al

Edge Al

less than 2 years 2 to 5 years 5 to 10 years obsolete before plateaumore than 10 years

The key challenges mentioned above cannot be addressed by using traditional operating models.

AI, combined with and empowered by advanced analytics, big data and virtualized computing power, will drive the automation and enhancement of network operations, enabling new operating models.

In July 2019, Gartner released a new technology maturity curve. At the peak of the curve, 12 technologies related to AI, such as AI PAAS, Auto ML (automatic machine learning), and others, have created high expectations about the adoption of AI. AI can handle massive data more effectively, select the relevant information more precisely, provide decision support, and automate human tasks.

Below is the technology maturity curve issued by Gartner in 2019, showing that AI technology is on the peak of inflated expectations, and, especially, that edge AI is about to be mature within 5 years.

1.4 Deploying AI is the trend in industry

Figure 15: 2019 new technology maturity curve from Gartner


419 778 1,359

3,9545,364

8,233

9,85311,158

2018

预测性维护

防范网络安全威胁

智能 CRM 系统

运维监控和管理

欺诈预防

客户服务 & 营销

客户体验提升管理

视频压缩

2019 2020 2021E 2022E 2023E 2024E 2025E

SoftwareHardwareService

38,38434,561

29,348

18,78213,829

4,7122,6761,412

2018 2019 2020 2021E 2022E 2023E 2024E 2025E

+52.12%

Omdia/Tractica a global leader in emerging technology market research, has conducted a research on nearly 300 real AI use cases in 30 fields and found that the telecom industry is particularly active in AI technology, being the largest potential AI market segment in the future. According to Omdia/Tractica's forecast, by 2025, the telecom industry will invest 36.7 billion US dollars globally in AI software, hardware, and services. The AI software market for the telecom industry will grow from $419 million in 2018 to $11.3 billion in 2025. It is expected that by 2025, telecom operators will heavily use AI for network operations, maintenance, monitoring, and management, which will account for 61% of AI expenditure in the telecom industry, as shown in Figure 16.

The gradual deployment of AI and big data technology will drastically increase the level of automation, improve service performance and quality, and reduce maintenance costs.

Source: Omdia/Tractica “Global AI revenue forecast 2018–2025”

Source: Omdia/Tractica “Global AI revenue forecast 2018–2025”

Figure 16: AI revenue forecast of telecom industry from Tractica

Figure 17: Global Telecom AI software revenue forecast by application scenario ：2018-2025

13 Future Operating Model for Network Operations

Future Operating Model for Network Operations2

Key capability requirements for next generation of operations and maintenance

• Introducing AI to assist decision-making• Applying of big data and AI • technology for prediction and prevention• Making decisions automatically

Closed-loop automation and intelligence

• Big data platform of operations and maintenance

• Ability to build network management platform and open platform

Platform and capacity opening ability

• Iterative development, fast response to business needs

Flexible business development capabilities

Meet the needs of diverse service , fast iteration

Predict and prevent, improve user experience

Introducing AI to assist decision-making

Fault diagnosis relies on expert experience, which takes a long time to deal with the group of obstacles and has low accuracy

Chimney OSS system, software change cycle is difficult to control, new business online cycle is long

Fragmented processes, segmented by areas, with a large amount of labor, resulting in a surge in costs

It's hard to deal with large-scale network with accumulated experience

New

Requirements

Challenges

Efficiency

Intelligent

Separation of operations and maintenance

system



Difficult to solve faults

Quality

In order to address the challenges described above, a new operating model for the network operations and maintenance processes and activities must be designed and built. This new operating model should include a new generation of operational capabilities enabling flexible business development, closed-loop automation, AI-driven processes, open platforms, and operational knowledge capitalization, as shown in the figure below.

Efficiency

Quality

Quality

Quality

Intelligent

Intelligent

Intelligent

Separation of operations and maintenance system






Difficult to solve faults

2.1 Requirements for the future operating model

Figure 18: Capability requirements of future operations and maintenance mode

• Expert experience assets for continuous iteration

• New operations and maintenance assets• Precipitated by artificial intelligence

operations and maintenance knowledge asset capability

14Future Operating Model for Network Operations

Main application of man-machine cooperation's mode in network operations and maintenance

• Manage network quality

• Failure prediction and prevention

• User behavior prediction

• Alarm compression• Fast troubleshooting• Quick service

opening

• Fault intelligent identification

• Intelligent fault diagnosis• Unknown problems found• Manage network quality• Failure prediction and

prevention• Recommended measures• Automation• Fault self-repair• Active service

Fault management1

• Order creation• Resource

allocation

Service open2

• Manpower optimization

• Intelligent scheduling

• Route planning

Exterior line management

3

Service experience4

• Network topology visible

• Real-time network monitoring

• Performance degradation identification

Network performance

5Network performance5

Service experience4Service open2

Fault management1

Exterior line management

Network deployment

Network planning

Product planning

3

Unstructured,changeable,

large in amount

General,predictable,rule-based

Temporary, indeterminate,forecast-based

Complexity of work

Structured,stable,

small quantity

Data complexity

Effectiveness driver

Efficiency driver

Innovation driver

Customer experience driver

Automation

Capability enhancement

Through the connection of knowledge, people and machines can work closely together. By exploring the unique advantages of artificial intelligence, we can solve problems and challenges that could not be fully addressed by traditional methods so far. The main capabilities of AI technology are:

Let's analyze the fault management process, which is one of the key processes to ensure normal network operations. In traditional operations, when network components fail, operation engineers handle the faults reactively according to their knowledge and experience and with the help of instructions summarized in knowledge management tools. Traditional fault management requires a lot of human experience and is consequently inefficient and expensive.

In the figure below, we classify network operations activities according to four main drivers:

These are all key drivers for moving to intelligent operations.

1 AI software has learning ability. It can analyze new input data, learn from them, and strengthen the AI model through continuous learning. By mastering the AI/ML training process, the accuracy of the AI models can be continuously improved.

3 AI is efficient. It can simulate human behavior to conduct repeatable work, improving production efficiency.

2 AI can handle and explore volumes of data and details where humans would hardly keep up with the pace.

Efficiency Customer Experienc Effectiveness Innovation

2.2 The role of AI in telecom network operations and maintenance

Figure 19: Application of artificial intelligence in network operations and maintenance


Alarm monitoring (Alarm management,Work order creation)

Typical applications

Value analysis

Service counter

First lineSecond line

Exterior line |Third line

BO

Monitor & assign customer initiated work orders to first / second line

Network

Third Line

Exterior Line Engineer

Network uptimeSLA compliance

Inventory / spare parts cost

Net recommendations / user experience

Data (Journal, Probe…)CST

Fault management (Work order management)

Fault & problemmanagement (Update the work ordermanagement)

A

Machine workflow

Knowledge asset

platformHuman

workflow

Failure prediction and prevention

Intelligent fault identification

Intelligent work order scheduling

Intelligent fault diagnosis

Automatic troubleshooting

A

B

C

D

E

C

B

ED

ALARMS TT

TT

TT

TT

Alarm

Exterior line costDoor-to-door service

Outsourcing cost

Customer loss rate

Perceptual analysis model

Strategy logic model

Control & operations API

Data model

Process assets

Report template assets

Interface display assets

System integration assets

TT

TT

WO

AI technology, combined with the experience of operations engineers, enables the set-up of intelligent and automatic monitoring and fault management systems.

The new AI-driven fault management systems collect real-time network management data and alarms, and perform data analysis, flexible filtering, classification, fault localization, and instant diagnosis on network faults with the support of well-trained AI algorithms. Fault location and root cause analysis then become more accurate, fast, and efficient.

Additionally, AI enables the prediction and prevention of faults, transforming this process from reactive to predictive and allowing network issues to be solved before end users notice anything. Continuous self-learning based on historical and new input data ensures a gradual and progressive improvement of the AI-driven fault management system's accuracy and reliability.

The picture below shows how AI is working in the fault management process in the intelligent operating model.

Figure 20: operations and maintenance mode of artificial intelligence in network fault management

As AI continues to mature and play an increasingly important role in operations, network operations management will be shifting from the traditional "person + process" model to the new "Human + Machine" Collaboration Model.

In the "Human + Machine" Collaboration Model, machines assist operations by applying AI and machine learning (ML) algorithms to large data sets collected by operations support tools and business systems, and simulating human thought processes and behavior, such as risk perception, pattern recognition, data analysis and interpretation, complex decision-making, and execution of elaborated tasks. The "Human + Machine" Collaboration Model provides operations management with AI and ML capabilities that free operational personnel from repetitive and routine tasks that machines can do even better than humans.

2.3 Future model of network operations: "Human + Machine" Collaboration Model


Role of human

Data is extracted into

knowledge

Role of machine

Design state: inject human experience

Operating state: machine self decision-making and self execution

Arrange

Perception Analysis Decision-making Execution

Multiple data sourcesStandard model

Preventive analysisAccurate identification

Fast decision-makingPrecise positioning

Accurate executionSelf healing closed loop

Design Train Feedback

• Status query API• Topological model• Engineering

parameter model

• Preventive prediction model

• Tuning logic• Troubleshooting process

• Network configuration API

• Network operations API

• Fault model • Fault tree• Diagnostic logic• Root cause analysis

model

Role of person:

The first generation operations and maintenance system:

Manual operations and maintenance

The second generation operations and maintenance system:Process automation

The third generation operations and maintenance system:

Man-machine coordination

Nature of data:

Role of machine:

Knowledge extraction and internalization

Decision making and execution

Knowledge extraction and internalization Knowledge extraction and injection

Information collection record

Decision-making and execution

Process automationInformation collection record

Process supervision

Original operations data Original operations data

Supplementary analysis data

AI training model

Supplementary analysis dataOriginal operations data

Arrange knowledge

Self-decision-making, self-executionProcess automation

Information collection record

Data Knowledge

Figure 21: The role and relationship of human, data and machine in the iterative development of operations and maintenance system

Figure 22: Diagrammatic sketch of operations and maintenance mode of new generation telecommunication network

Under the new operating model, new technologies will greatly empower the traditional operations teams. The human's unique experience and judgment skills are being transformed into data and then injected into the knowledge base. The machine learning ability will form the closed loop of telco's self-decision and self-execution capabilities. As a result, the convergence of human insight and machine learning will offer unlimited possibilities to network operations.

Below, the figure shows how the role of person and machine is evolving through different development stages of the intelligent operations system transformation.

In the future operating model, the roles of humans and machines will change significantly. At the design stage of the systems and processes lifecycle, people will transfer their knowledge and experience on complex operational domains to machines through the ML training process. Machine learning will facilitate the continuous advancement of computing through exposure to new scenarios, testing, and adaptation, while employing patterns and trend detection for improved decisions in subsequent (though not identical) situations.

The figure below shows how machines and humans are working and interacting with each other, and how is this interaction enabled by programming data.


Reduce cost Improve quality

Number of passive complaints from

customers

Opening experience

Service open

Serviceexperience

Exterior linemanagement

Network assuranceNetworkperformance

Networkdeployment

Network planning

Product planning

User experience

Accurate and timely

performance warning

Accurate and timely

complaint warning

Accurate and timely

network monitoring

Order omission

probability

MTTR

Performance degradation identification

User behaviorprediction

Network topology visibility

Fault reporting

experience

Waiting time for opening

Number of fault reports

Work order

quantity

Times of log in

Customer-service labor time for

reporting faults

Labor time for fault location and

diagnosis

Labor time for manual dispatch and

documentation

Reduce operations andmaintenance expenditure

More efficient operations

and maintenance

operations and maintenance

with higher quality

More intelligent operations

and maintenance

Labor time for troubleshooting

Labor time for opening configuration

Efficiency of system monitoring and identification

Scale of operations

and maintenance

eam

Data complexity

Work complexity

Efficiency of operations

and maintenance

team

Number of invalid Group Fault repetitions

Big data and AI technology are used to predict and prevent network failures, greatly reducing the number of failures and complaints and improving the perceived quality of the network.

Additionally, the new AI-driven operating model enables the control and management of the growing complexity and diversification of the network environments.

At the same time, the conflict between complexity, maintenance costs, number of people, and other operational workloads will be mitigated, alleviating the pressure on cost-optimization programs.

2.4 Business benefits of the "Human + Machine" Collaboration Model

Figure 23: Value analysis of intelligent operations and maintenance

Cost Quality

Transformation

External business transformation

Upgrade from 3G/4G to 5G service

New business of industry

Internet of Things

Real time digital service

Enabling business ecology

Integration of consumer

and family business

Cloud network and telecom cloud service

End to end business insight

Upgrading talent structure

Internal capability transformation


In network assurance, service provisioning, and 3rd party

maintenance service, repetitive and ineffective manual work can be

reduced through automation and AI, leading to a reduction in team

size and great improvement in operational efficiency.

By moving from reactive to predictive and defensive maintenance, we can

discover and solve hidden issues, minimize network interruptions and user complaints, and boost network availability, improving service and customer experience at the same

time.

The transformation to the new operating model involves platform, processes, and workforce, making

network services no longer a bottleneck of business development,

and supporting the improvement of digital services, business insights,

talent management, etc.

Cost Reduction Quality Improvement Total Transformation

In summary, the business benefits brought by the "Human + Machine" Collaboration Model are:

Defined, bounded, and strictly controlled procedures

Self learning, self-rule, borderless automation Fixed automation

(based on rule)Evolution automation (based on judgment)

Bases Databases Standardization of activities API utilization

The first step of the operating model transformation is to build a basic intelligent platform, and then implement simple, intelligent applications on the platform, focusing on addressing relevant operational issues and pain points. It starts from a single-point application scenario to achieve a single-point intelligence, such as implementing automatic application inspections, intelligent early warning indicators, etc., and initially establishes a data system to import knowledge of single-point scenarios, realizing the series intelligence，and completing the data system based on perception, analysis, decision-making, execution, and intention. Finally, the whole scenario becomes a highly integrated intelligence, achieving closed-loop intelligent operations. The operational personnel can gradually shift their focus from tedious operational activities to exploring needs, defining scenarios, and performing other key tasks.

Figure 24: Evolution path of operations and maintenance mode

2.5 The evolution path to the "Human + Machine" Collaboration Model

The stages leading to the adoption of a full "Human + Machine" Collaboration Model are the following:

The figure below shows this transformation path:

single-point intelligence high intelligenceseries intelligence

Single-point intelligence

Series intelligence

Highly intelligent

Focus on the top scenario, identify the capability framework according to perception, analysis, decision-making, execution, and intention, and gradually establish the operations and maintenance data system, focusing on the pain points, to solve the problems.

The operations and maintenance data system based on the five dimensions of perception, analysis, decision-making, execution, and intention is completed. The main operations and maintenance scenarios are flow-based and intervention-free, and business problems are quickly transformed into data problems.

All scenarios can be covered, and intelligent analysis can be adjusted among operations and maintenance cost, quality and efficiency to support business decision-making, service content, and mode upgrading.

19 Intelligent Operations of HUAWEI

Intelligent Operations of HUAWEI3

HUAWEI has provided more than 580,000 instances of technical support, over 10,000 instances of problem positioning, and more than 130,000 instances of network major operations support for more than 1,700 networks around the world. Based on this rich network operations experience and our innovation practices, HUAWEI is proposing an AI-driven human-machine cooperative intelligent operations solution.

At the design stage, knowledge of skilled people is captured and codified in the knowledge management platform. Combining this knowledge with comprehensive input data, ML training, and process design thinking, we implement into our platform the repetitive and deterministic tasks previously performed by operations people, and we continuously improve them. Through such iteration, more and more jobs are gradually handed over to machines, improving the accuracy of the operations and maintenance services. In production, the machines run autonomously. In addition to scheduled operations and maintenance tasks, we implement predictive and self-healing capabilities into our platform using big data and AI.

The human-machine collaboration operating model will leverage and value existing operations systems and tools, deploying the knowledge management platform on top of them and integrating existing processes, interfaces, and needs.

The human-machine collaborative operating model is not a substitute for human operations, but an enabler for people to create more value with the assistance of the machine. As shown in Figure 24, by upgrading the skills of current operations personnel, such as front office (FO) engineers, back office (BO) engineers, 3rd-party engineers, etc., they are transformed into network strategy engineers, application orchestration engineers, data engineers, and other new operations roles, letting people playing key roles in solution design, exception handling, and decision-making.

3.1 Design principles of HUAWEI's intelligent operations and maintenance service

Integrated human-machine cooperative operations and maintenance

20Intelligent Operations of HUAWEI

Operating

state flow

Design

state flow

Predictable

Perception

Train

Decision-making

Arrange

Execution

Feedback

Design

Machine

Datafeedback

Knowledge injection

operations and maintenance knowledge asset platform

Person

Analysis

Self healing

Network strategy EngineerData analysis EngineerRealize data collection, conversion, cleaning and aggregation analysis

App Arrangement EngineerArrange process logic and design manual

intervention pointsDesign analysis logic and execution strategy

Perceptual analysis model

Strategy logic model

Control and operate API

Model assets for reports Data model

System integration and

adaptation

Interface display control

widgetProcess assets

Perception Analysis Model

Process Assets

Common Capabilities

Reporting Template Assets Presentation Interface Assets System Integration Assets

Policy Logic Model Control API Data Model

Fault Models

Prevention & Prediction Models

Root-Cause Analysis Models

Fault Processing Process

Change Mangement Process

Patrolling Process

Diagnostic Logic

Optimization Logic

Poor-Quality Location Report

Network Capacity Reports

Workforce Mangaement Reports

Logic Tree

Network Configuration API

Status Inquiry API

Network Ops API

GIS Heat Map

Topology Map

Event Monitoring

Third-Party Performance Data Collection

Third-Party Work Order Integration

Third-Party CRM Integration

Topology Models

Leased Line Models

Engineering Parameter Models

Generic Presentation Controls

AI Algorithms and Models

General Data Integration

General Data Processing

The HUAWEI Intelligent Ops platform aims to support the next generation of operations powered by machine and knowledge. Using AI-based automation, HUAWEI's platform continuously absorbs operational knowledge and transforms concrete use cases into componentized services. HUAWEI Intelligent Ops includes three types of offerings:

As shown in Figure 25, the knowledge assets include the following categories: perception analysis model, policy logic model, API control & operations, data model, process assets, reporting template assets, presentation interface assets, and system integration assets.

Service offerings

Figure 25：Huawei’s hyper-integrated operations and maintenance concept of man-machine coordination

Figure 26: Classification display of knowledge assets


“Use Case-Based” Assets

Standardize

Network Technology Offerings

FBB MBB 5G & NFV Professional Services

EcosystemOfferings

Open Capabilities

Basic Functions

Design Studio (DevOps-friendly，orchestration-friendly）

Mapped to conventional OSS: Faults, work orders, performance management, etc.

“Componentized” Assets

Standard Reusable Asset Library (OOTB)Project-Based Asset Library

(Ecosystem)

New Capabilities

Conventional Functions:

FBB-Home Internet

MBB-Wireless Fault Automation

FBB…

Control

NetworkOps

NetworkAnalysis

DataModels

Process Assets

Reset

Reactivation

Parametric Tuning

Linked Data

Fault Tree

Traffic Predication

Event Model

Topology

EngineeringParameters

Scheduling Process

Patrolling Process

Spare Parts Management

GUI

I/O

Data Processing

5G…MBB…

In addition, HUAWEI offers professional services, such as deployment and consulting services, to accelerate our clients' success.

provide standardized use case–based offerings powered by advanced and innovative technology (AI, Big Data,

Virtualization, IoT, etc.). Leveraging our experience on network operations in

160+ telcos, HUAWEI offers professional operations solutions for different

use cases, such as wireless network intelligent operations (MBB), fixed

network intelligent operations (FBB), 5G and NFV intelligent operations, etc.

offer "componentized" services that encapsulate capabilities into APIs and

other programmable components and toolsets, including AI Training

Center (integrated toolset that combines data training, models, and algorithms), Design Studio (automated process packages,

operational APIs, model packages, and data service packages), and Data Governance Services (data collection,

cleansing, and modeling).

provide OSS functions, including fault management, performance

management, and workforce management. These services are now

virtualized and integrated with the centralized data platform, knowledge management services, and run-time services performing the perception / analysis / decision / execution cycle.

Network Technology Services Ecosystem Services Platform Services

Figure 27： Huawei's intelligent operations and maintenance works

5G B2B Business

Assurance

Network GDE

FM

Perception Analysis Decision Execution

Data Collection + Transformation + ModelingData Governance

Data Service + Model Assets + Algorithm Assets

AI TrainingComponentized Automated Processes + Control API + Model Service + Data Service

Design Studio +

PM SDM WFM APSEMS

YourTelcoPerformance

Reports

South Africa GIS Heat

Map

Home Internet Intelligent

Ops

Leased Line Intelligent

Ops

Home Internet Intelligent Activation

Automated Mobile Fault Management

Automated Mobile

Performance Ops

Backbone Network Intelligent

Ops

5G Intelligent Assurance

Adoption Service

Development Enablement

Service

Customized Services

(AI Enablement，

Data Services)

NFV Intelligent Assurance

5G B2B Business

Assurance

Orchestration Knowledge Network Ops Knowledge Assets

Platform Offerings


As-Is Workforce Model

Operational Use Cases

“0” Touch

Complaints / Voice of Customers

Data Training Center

• Data ETL• Data modeling• Feature engineering• AI model training• AI model

optimization• Data security

• Event model adaption and instantiation• Application orchestration and development• Rules and policies formulation

• Uncertainty decisions• Emergency intervention• Smart outcome labeling• Irreplaceable work

(hardware installation, complaint handling, etc.)

ML/DL Training Platform

Data ScientistsData Engineers

Data model orchestrationProcess automation orchestrationModels

/ Rules / Policies

Training Data

Reusable Model

Real-time / batch data

Automated Closed Loop

Manual On-Site operations

Automated Work Orders

Event instantiation / analysis / policy /

automation configuration

Policy & Automation

Analytics

Big Data + AI

SO

Unified Data CollectionSDM

DIM

NOC（NPM)/SOC

Field Engineers CC

operational Orchestration Studio

UI interface orchestrationAutomated use case orchestration

operations Data Lake

Event FlowInter-System InteractionsHuman-Machine Interactions

Transformed into new posts

Network Policy Engineers

BO FO

Field Engineers

Application Orchestration Engineers

Data Analytics Engineers

DispatchingChange Execution

Remote RepairPatrollingMonitoring

Whitebox Contractor Management

Cost-Saving Sharing Model

Remote Fault Processing

Retained with changed R&R

Substituted by machine

Substituted by contractors

Transformation through Intelligent Ops

Figure 28： Logic structure diagram of intelligent operations and maintenance

Figure 29: Personnel transformation

HUAWEI Intelligent Ops transforms not only the systems, but also the operations culture, processes, and working style.

In the new operating model (Figure 27), data scientists and data engineers in the training center collect data from the operations data lake for data cleansing, data management, and modeling. Engineers in the design center build data models, design the user interface, and create policies. The NOC (NPM), SOC, and local operations teams will conduct emergency interventions and make decisions in escalated ad-hoc situations.

Traditional back office (BO), front office (FO) and subcontractors will gradually be adapted to fit into the new organization. As shown in Figure 28, around 20% of the positions will be transformed into Network Policy Engineers, Application Orchestration Engineers, and Data Analytics Engineers. About 30% of the original workforce would be replaced by machines handling automatic monitoring and device inspection. Current operations personnel can be retrained and reskilled to cover design management, supervision, and governance roles, giving them the opportunity to grow professionally into new career paths.

Transformation of operations and maintenance model

Orchestration Design Center

NOC(NPM) / SOC / Local Ops Team

Network

BO Engineers

20%

20%

30%

30%

FO Engineer

Field Engineers

• Change requests• Major incident handling

• SLA monitoring• Automated task monitoring

• Contractor management


The domain knowledge and expertise will continue to be harvested into the platform. It will be "digitized" and "capitalized" so that the platform can become "smarter." Knowledge mainly comes from the technical and operational experiences and insights, which is captured from Network Policy Engineers, Application Orchestration Engineers, and Data Analytics Engineers. They will take the knowledge assets from different FBB, MBB, and 5G / NFV use cases (standard reusable libraries, performance reports, heat maps, etc.) and package them into components in the orchestration center, then customize them with real-world insights, human experiences and professional reviews. Driven by AI and ML, the knowledge-based platform will be able to power the machine's run-time analysis-decision-execution-perception cycle.

In HUAWEI's Intelligent Ops platform, we have overcome the siloed architecture. We have gradually decoupled the existing network management applications and transformed them into a layered, service-oriented architecture. Using a unified middle layer, data from different business lines can be pooled and shared. We provide a centralized AI training and application platform to better respond to the market demands of the telecom industry.

In the figure below, we show the components of the Intelligent Ops architecture.

Figure 30: Knowledge transformation

Project-Based Asset Library (Ecosystem)

Standard Reusable Asset Library (OOTB)

“Componentized” Assets

“Use Case-Based” Assets New Capabilities

YourTelcoPerformance Reports

FBB-Home Internet

FBB…

MBB-WirelessFault Automation

MBB…

5G 2B Business Assurance

5G…South Africa GIS Heat

Map

Orchestration Knowledge Network Ops Knowledge Assets

Design Studio (DevOps-friendly，orchestration-friendly）

Control I/O

Data ProcessingGUI

Reset

Reactivation

Parametric Tuning

Linked Data

Fault Tree

Traffic Predication

Event Model

Topology

EngineeringParameters

Scheduling Process

Patrolling Process

Spare Parts Management

Network Ops Network Analysis Data Models

3.2 Solution architecture


operational Use Cases & Assets

Intelligent Ops Runtime

FBB Intelligent Ops

Cloud-BasedAI Training

Design Studio

Knowledge Assets

operational Service

Continuous renewal of algorithms & models

Continuous renewal of event handling policies

and models

AI & Algorithm Service

Policy Automation Service

DECISION

Dynamic Resource Service

Automated Offering Activation Service

EXECUTION

operational Analysis Service

ANALYSIS

Unified Data ServicePERCEPTION

MBB Intelligent Ops NFV & 5G Intelligent Ops

ExecuteEvent Trigger

Intelligent Ops Assisted Service

Monitoring

• Work orders• operational intelligence• Dashboard and cockpit

• Alerts• Performation• Offerings

Diagnose

10

5 4 6

7321

8

Unified Data Collection & Command Dispatching Service9

Figure 31: Framework of Intelligent Ops

1 Monitoring (alarm / performance / service status): Basic support services, including alarms and metrics monitoring as well as comprehensive service status and performance monitoring. This component provides alarm correlation analysis, RCA, KPI aggregation calculation and layout, and dashboard report functions.

2 Operations Analysis Service: Based on the original or correlated alarms, as well as KPI thresholds, this module comprehensively analyzes and identifies operational events, including fault events, service degradation events, operational risks, etc., establishing an analytical mechanism for use case–based faults.

3 Dynamic Resource Service: Unified resource management module, supporting dynamic modeling, unified management of business and data correlation, and quick search and presentation of topology.

4 Policy Automation Service: Problem localization, using issue-tree and event-analysis tools to achieve self-healing (executed through automatic triggered actions).

5 AI & Algorithm Service: This component provides AI algorithms and analytics services, such as alarm correlation compression, trend prediction, KPI detection, etc.

6 Intelligent Operations Supporting Services: This module provides work orders and external process management, restricted access to mobile phones, and screen monitoring capabilities.

7 Automatic Business Opening Service: Automatic execution of various service activation requests.

8 Unified Data Service: Big data analytics service, providing distributed computing and data orchestration functionalities.

9 Unified Data Collection and Instruction Adaptation Service: Unified southbound command layer, interfacing with network management system, adapting the interface with the network admin system, and implementing data collection and command dispatching.

10 Operations Use Case & Assets: This component provides operations management solution assets based on fixed networks, mobile networks, and 5G / NFV distilled from 160+ operation engagements.


3.3 Core capabilities

With extensive project experience and a rich technology portfolio, HUAWEI is a reliable and highly competitive player in the telecom industry with a deep understanding of MBB / FBB / NFV and 5G use cases.

Domain expertise in telecom use cases

HUAWEI has the know-how to significantly reduce the fault rate through prediction of copper-access quality issues and remote optimization algorithms. The same applies to accurate prediction and isolation of poor optical-access quality and broken fiber. Intelligent Ops can effectively reduce home visits and on-site work through the interception of correlated tickets and AI-driven fault diagnosis.

Through the automatic analysis of work orders (involving automatic processing and lean WO dispatching), AI-based fault identification (involving cross-domain topology restoration, general fault extraction, and cross-domain fault identification) and AI-based fault diagnosis (accurate diagnosis, automatic fault repair, and root cause recommendations), we ensure a significant improvement of the FO and BO efficiency. In addition, we reduce the number of invalid orders and network service interruptions.

HUAWEI provides effective SLA quality visualization, active SLA management, SLA degradation prediction, policy-based self-healing, leased line availability improvement, customer complaints reduction, and SLA violation rate reduction. Based on real-time service performance monitoring, Intelligent Ops can precisely locate and isolate the faults, reducing the corresponding repair time (MTTR).

we enable the improvement of metrics such as stations per person, diagnostic accuracy, and MTTR of performance degradation by using our performance degradation order automation, performance degradation identification (HUAWEI + MV wireless performance monitoring model and dynamic threshold algorithm based on machine learning), and performance degradation isolation (fault isolation expertise, clustering analysis algorithm, etc.) solutions.

HUAWEI has a lightweight solution for home Internet service activation. It covers mainstream access scenarios such as FTTH, WTTH, FWA, etc., to meet the demands of small- and medium-sized telcos for quick service activation, at very low cost. For 5G, HUAWEI offers FWA SA activation and FTTH / 5G FWA intelligent integrated activation solutions.

we use our core network risk prediction (data feature extraction for fault prediction, KPI-based anomaly prediction, and CHR- and log-based anomaly detection) and core network risk and problem location and isolation (issue tree, fault data clustering and correlation analysis, cross-domain fault isolation, cross-domain network topology restoration, and AI-based assisted isolation) solutions to continuously optimize the core network services. We improve redundancy coverage and reduce the MTTR. Moreover, Intelligent Ops can decrease unscheduled service interruptions caused by changes related to anomalies in core network.

Intelligent Ops Home Internet Service

Wireless Network Fault Management Automation

Intelligent Ops Leased Line Service

Wireless Network Performance Management Automation

Home Internet Service Automatic Activation

Intelligent Ops of Core Network

FBB

MBB


Real-time updates of dynamic topology. Minute-level automatic restoration of service path. MBB (Mobile Broadband) network quality visualization.

Wireless event diagnosis and cloud-based core network cross-layer problem solving (minute-level).

NR equipment health check. VoNR (Voice/Video over New Radio), and data redundancy prediction and prevention. NFV health check.

Minute-level updates of slicing topology. Slicing quality and SLA visualization & diagnosis.

Visualization and Management

Intelligent 5G Fault Management

5G Network Service Prediction & Fault

Prevention

Cross-Domain Network Slicing Visualization &

Management

5G & NFV

Change in Use Cases Perception Analysis Decision

Execution

AI

Performance Alerts

Complaints

Power Environment

Device Alerts

Log

From Access Network to Bearer Network

MBB Automated Event Analysis Automated Policies

Alerts reduced by 90%+

From alert-oriented to business impact

-oriented

60% of events need positioning, of which 80% can be traced by fault tree and rules.

(Decision Tree & Rules)

Automated Work Order Dispatching

Manual Repair

Automated Closed Loop

Legacy Network NFV Cloud5G SDN

• Fault event• Worsening events• Risk events

Automated Work Order

Creation

Business Impact &

Event Zoning15% Risk forecast

Root causes located 90%

RedundancyCompression

Event Identification

Automated Diagnosis Policy DecisionsAutomated Diagnosis and Decision

Predicative Algorithm

Issue ZoningFrom Complaints to Predication

FBB

From Single Domain to End-to-End

NFV&5G

From physical to logicalFrom generation to

recovery

自动化API目录

Log/Status/Config/info

• 诊断 API 库

Reboot/Reset/Migrate/Config/…• 自愈 API 库

HUAWEI supports predictive fault management by unifying AI capabilities and intelligent data governance and services.

The figure below shows the HUAWEI "Use Case–Based" Intelligent Ops.

AI and data analytics

As shown in the Figure 32, HUAWEI infuses AI capabilities into its operations, algorithms, and offerings. With the Intelligent Ops solution, AI algorithms are packaged into components to make them easy to reuse.

Unified AI Algorithms

Figure 32: HUAWEI "Use Case–Based" Intelligent Ops

Human-Assisted

Business Orchestration & Command

SystemSelf-

Healing


Before Event: Prediction & Prevention After Event: Diagnosis and Recovery

Collection Framework

Network Worsening Model

AI Algorithm

Fault Tree DeductionUser Model

Resource Topology

Event IP Layer Model

Physical Layer Model

Application Layer Model

Link Layer ModelSample Repository

Flow Processing

Data Cleansing

Customized Data Probes

EMS General Probes

Topology Template Probes

User Feature Data

Probe Visualization

Output Standardization

Protocol Management

Data Redundancy Elimination

KPI Aggregation

Feature Extraction

Events Resources

Policies & Rules

Resources CHR

Alerts PerformanceData Security

Meta Data

Data Standardization

Data Quality

AI Model Training

Multi-Dimensional

Linkage

Execution Commands

Fault Tree

Feature Model

Decision Tree

Decision Data

Data Analysis

Cross-Domain Model

Data Governance Tools

Basic Model

Storage & Pre-

Processing

Resource Topology

Fault Tree

Physical Layer Topology

Business Topology

AlertFault Predication Completeness : 50% Root Cause

Identification Coverage: >85%Fault Prediction

Accuracy: 90%

Data Services: Data processing and analytics for business Data Governance: Standardization for business

MBB Ops

• VoLTE anomaly detection and cluster analysis (redundancies)

• EPC traffic prediction• Cross-domain alert correlation

(wireless+microwave, wireless+microwave+IPRAN, wireless+PTN)

• Wireless access KPI anomaly detection (prevention)

FBB Ops

• Home Internet copper fault predication (to reduce complaints)

• Home Internet optic access quality worsening, broken fiber detection (to reduce complaints)

• Home Internet user behavior analysis for complaint reporting

• Corporate service anomaly detection• RPA for customer services

5G & NFV Ops

• NFV fault detection and analysis (tor redundancies), 5G fault detection and analysis

Business Enablement

Use Case-Based AI Models

Generic AI Algorithms

• Images• Voice• Semantics

• Sub-optimal hardware• Connection worsening• Capacity and traffic

• Time series anomaly detection• Event data (alerts, log, trace, CHR) anomaly detection

• Alert correlation• Root cause recommendation

• Complaint analysis

• Fault tree• Knowledge map

Feature Engineering Model Management Online Training AutoML Active Learning

Build use case-based models Generalize AI algorithms

Figure 33: Data enabled fault prediction and management

Figure 34: Unified AI algorithm capability

As shown in the Figure 33, high-quality data is the foundation of intelligent operations. Thanks to HUAWEI's data collection framework, we broke the siloes and now manage a common data lake. Using standardized data models for multiple telco services, Intelligent Ops has laid an effective analytics foundation for AI training, enabling applications such as fault prediction and prevention, automated fault management, root cause recognition, etc.

Data governance and service

Human-Machine Interface Prediction Detection Decision Knowledge


Intelligent Ops Capability Center

Intelligent Ops Knowledge Assets

Provisioning Internalization

Process Automation

Presentation Layer

Logic LayerD

ata Layer

Process forms & controls

Process logic &

automation rules

Managed objects model

Ops Asset Library

Real-time Monitoring Controls

Network Event Model

Data Collection Probes

Ops Process Reports

网元操作 API

Network ModelSecurity Management Maintenance

Data aggregation &

linkage

Realtime AI analysis

Managed objects model (service, site, device, connection, board, ports …)

Raw data collection (alert\performance\log\resource …)

Data extraction logic

Non-real-time AI analysis

Automated use case API

Automated atomic API

Activation use case API

Activation atomic API

Monitoring dashboard & operational orchestration

Styles and data-drilling

logic

Activation form & control O

psD

evelopment

Automated operations

Reporting & Periodical

Analysis

Service Activation

Use Case-Based App

1

Use Case-Based App

2

Use Case-Based App

N

…

Monitoring & Real-time Analysis

Idea App

Ecosystem App

Testing / Deployment

Configuration Service

Design data model

Understand business needs

UI Design

Telco Ops Specialist

Encapsulate complex issues into use case-based solutions

Open capabilities in orchestration center for telco to build its own applications

Integration Service Specialist

Design Studio

Telco or ISV Developers

Figure 35：operations ecosystem transformation

HUAWEI aims to simplify the operations of our clients by encapsulating the rules, models, and AI algorithms distilled from a great number of successful real-world engagements into a unified Intelligent Ops platform. We give our customers and partners access to this knowledge so that everyone in our business ecosystem can develop their skills by leveraging that experience.

OWS (Operation Web Services) is the enabling platform for HUAWEI Intelligent Ops, with big data and AI capabilities, which can be quickly deployed in any telecom operations environment. It is designed to adapt quickly to the evolution of technology and telecom services, enabling the telco's engineering team to develop customized applications through open APIs. Based on the microservice architecture, OWS decouples IT services from business services. Its unique service orchestration ability will enable telco engineers worldwide to address their specific operational needs. HUAWEI is building a global developers' ecosystem where telco and other HUAWEI partners are able to develop applications in Design Studio, then easily deploy them into production (the whole system is shown in the figure below).

Operational ecosystem

Orchestration Center AI Training


As the foundation of intelligent operations, the data platform is one of the core components of the pilot application. We will integrate data management modules to collect, analyze, calculate, and store operational data, define a uniform metrics system, implement data importing and extraction procedures, and start storing and curating data for future intelligent operations’ needs.

We recommend telcos evaluate business needs and start from the most urgent use case. While we bring the intelligent operations application online for the use case, we will also build the essential platform capabilities at the same time, including the time series data system, anomaly discovery, traffic abnormality alerts, database monitoring, network log analysis, etc. Starting from a single pilot use case, a telco can gradually build a platform for future applications throughout all service lines.

Building Capabilities through Use Case Pilot

Data Platform

5G Intelligent OpsApplication

Layer

5G Fault Center 5G Performance Center

Prediction & Prevention Center

Policy Automation CenterDigital operations

Resource Management

Network Optimization Fault Management

Performance Management

Platform Layer

5G NFV

Corporate ServicesHome

Internet

3G / 4G

2G

Network Layer

Data Import

Intelligent operations cannot be achieved overnight. It can coexist with legacy OSS in the short term and then take over gradually. A telco can select the use case–based entry points with the greatest business impact, based on its own business needs and pain points, import operational data from the existing OSS system, and launch applications such as intelligent anomaly alerts, fault discovery and recovery, and smart inspections. The telco can protect its investment in the existing OSS, find a use case to introduce intelligent operations, bring it live, and then reinforce and replicate the success to the rest of the operations, eventually taking over the legacy OSS.

HUAWEI's intelligent operations-transformation plan is based on the principle of protecting operators' existing investment and supporting the smooth evolution of operators from the traditional operating model to the intelligent operating model of human-computer cooperation. The roadmap can be divided into three phases:

Phase 1: Single-Point Adoption. Evaluate the status quo and select a key application for the use case that demands new operational solutions and delivers the greatest impact. Start with a specific service scenario (e.g., 5G fault recovery), develop the corresponding application in the HUAWEI Intelligent Ops platform by adding an application overlay on top of the legacy OSS and a data layer to support perception, analysis, decision, and execution. Integrate the legacy OSS as a data source for intelligent monitoring, abnormality alerts, fault discovery and analysis, root cause analysis, and self-healing. Take this as an opportunity for building platform and data governance for fault and performance monitoring. The detailed transformation process is shown in the figure below.

3.4 Transformation plan to HUAWEI's intelligent operations platform

Figure 36: Phase 1

Intelligence Ops Applications

Conventional OSS8 Intelligence Ops Platform


Application Layer

Platform Layer

Network Layer

Home Internet Intelligent Ops

Corporate Intelligent Ops

Wireless Intelligent Ops

5GIntelligent Ops

Use Case-Based Intelligent Ops

Open Ecosystem & Programmable Ops

Open Interface for Capabilities

Application Layer

5GFault Center

Policy Automation CenterDigital operations

Resource Management

Network Optimization

FaultManagement

Performance Management System Platform

Platform Layer

Corporate Services

Home Internet

3G / 4G5G NFV

2G

Network Layer

Phase 2: Clustered Use Cases. A telco can extend the success of a single Intelligent Ops application to other use cases. In the application layer, we can replicate it to more value-based use cases (e.g., home Internet, corporate services, wireless, etc.) In the platform layer, we recommend a plan to bring Intelligent Ops modules online such as process center, orchestration center, and resource centers. The legacy OSS system functions will gradually be replaced in an orderly manner. The business value of the new intelligent operations will be unlocked through the process, and the telco will be able to capture new value in user experience improvements and cost savings in workforce and other operational resources. The detailed transformation process is shown in the figure below.

Phase 3: Highly intelligent operations and maintenance. In this phase, Intelligent Ops are now enabling all aspects of network operations, enabling Autonomous Driving Networks (ADN) with smart decision-making and automated closed loop for control and data feedback. Intelligent Ops have gradually replaced the legacy OSS for all service use cases. In the meantime, the telco will be able to reorganize its core operations workforce, moving most of it to design, training, supervision, and governance roles and responsibilities—as the most time-consuming operational tasks have been taken over by the machine and by the application orchestration center's engineering team. At this point, the full potential of the human+machine platform will be unlocked for the greatest business value impact.

The detailed transformation process is shown in the figure below.

Figure 37: Phase 2

Figure 38: Phase 3

Unified Ops Applications & Portal

Unified Ops Applications & Portal

OWS

API Gateway

OWSConventional OSS

API Gateway

5GPerformance

Center

Ops Process Center

Home Internet Intelligent Ops

Corporate Intelligent Ops

Policy Automation Center

5GFault Center

5GPerformance Center

CorporateServices

NFVHome

Internet 2G

3G / 4G5G

Predication and Prevention Center

System Platform

Ops Process Center Programming Center Resource Center

Wireless Intelligent Ops

5GIntelligent Ops

Use Case-Based Intelligent Ops

Open Ecosystem & Programmable Ops

Open Interface for Capabilities

Data Import

Planned Replacement

Programming Center

Resource Center

Predication and Prevention

Center

31 Success Stories

4.1 Large operator in China: Automatic fault management

Success Stories4

• Siloed OSS, data dispersion

» OSS tools are separated from each other with poor correlation of resources and data, and there is no end-to-end operational capability.

• There are many repeated invalid work orders and the proportion of invalid on-site visits is high

» 28.4% of the work orders do not have information regarding fault location, resulting in a large number of invalid on-site records.

» Limited accurate fault positioning and delimitation methods, abundant incorrect manual operations for troubleshooting，and a great deal of invalid requests for help together lead to low processing efficiency.

• The average fault recovery time (MTTR) is long

» It is difficult to communicate across different departments, and the average fault recovery time is 1,125 minutes.

• Maintenance depends on manpower, which is low skilled and high cost

» As the network scale becomes increasingly larger, more and more network resources are maintained, so it is difficult to rely on human labor.

» Decentralized troubleshooting methods and experience in cities cannot be shared. Most maintenance personnel are of high mobility, and generally lack accurate troubleshooting capabilities.

Operations & Maintenance Challenges

After many years of network construction, along with new technology development, a large telecommunications operator group in China has brought in many OSSs. However, traditional siloed OSSs have high maintenance costs and cannot be managed across domains and vendors. New services need to be integrated with multiple systems for online service deployment, so it is difficult to support the rapid launch of new business. As 5G, loT, and industrial digitization have increased network complexity, and multigeneration technologies have coexisted for a long time, delimiting and positioning network faults accurately have become more difficult and urgently needed.

Case Background

32Success Stories

• Reduction of the number of work orders and the client's operating cost.

» In one province, work order dispatches are now reduced by 2000+ per month on average, cutting down the cost of on-site labor. The annual cost has decreased by 2.4 million.

• Reduction of fault recovery time (MTTR) and improvement of operational efficiency.

» MTTR has been reduced from 1,125 minutes to 15 minutes and fault analysis efficiency increased by 75%. The accuracy of root cause locations reached 85%, considerably improving operational efficiency.

• Reduction of fault recovery time (MTTR) and improvement of operational efficiency.

» Delivered more than 20 fault diagnosis rules; covered 30 types of fault scenarios; dealt with 6800+ daily alarms; triggered 2500+ instances of daily automatic diagnoses with 75%+ success rate, reducing the dependence on personnel skills.

Benefits to client

Through building an intelligent operations platform, HUAWEI helped the group to connect the OSSs of different provinces and integrate the data sources. By using big data analysis and AI algorithm technology, HUAWEI helped them mine huge amounts of web data and expert experience, and continuously evolve according to each scenario's needs, with applications like the intelligent troubleshooting center and intelligent inspection center assisting the operators reduce the network risk, comprehensively improve maintenance quality, strengthen technical support for smaller cities, reduce the dependence on staff skills, and realize the transformation of the intelligent operations.

HUAWEI Solutions

Applicationlayer

Networklayer

Platformlayer

Figure 39: Fault automation solution architecture of a group

Three centers of the group

Intelligent operations and maintenance platformTraditional OSS

Intelligent service opening center

Intelligent network troubleshooting center （3G/4G,5G）

Electronic operations and maintenance

Strategy Automation Center

ArrangementCenter

Data Integration

Evolve asneeded Prediction and

prevention Center

Enterprise customer

2G

3G/4G5G NFV

Home broadband

FaultCenter

operations and maintenance process center

Central Platform

Resource Center

Performance CenterNetwork

optimization

Resource management

Fault management

Performance management

Intelligent inspection center(Cloud network, VoLTE)

33 Success Stories

The business of a large telecommunications operator in Southeast Asia includes fixed and mobile networks, with 2 million fixed network users and 65 million mobile network users. As the scale of the network continues to expand, the user perception of the entire network is poor. Under fierce competition from its competitor G operator, the number of users has dropped, and a modern operations system is urgently needed to enhance business agility, improve network quality, user experience, as well as operational efficiency.

On one hand, HUAWEI helps the operator build a unified intelligent operations platform. As shown in the following figure, it integrates the existing OSS systems, unifies data collection platforms, unifies operational processes, and provides an independent automation strategy engine based on the intelligent operations platform, making it possible for multiple network events to drive closed-loop automation. Facing future business needs, it provides alarm, performance, log, and other event-driven automation; builds an AI training platform; provides AI training capabilities based on massive operational data; and realizes automatic system learning.

Case Background

HUAWEI Solutions

Scenario-based operations and maintenance applications

Intelligent operations and maintenance platform

Existing OSS system

Customer complaint handling

Monitoring and management

Portal

Intelligent operations and maintenance auxiliary services

Surveillance

AI & algorithm service

TTS

Mobile network Hybrid cloudFixed net NFVSDNIT

FM PM IM T&D …

Unified data service

Unified data collection and command adaptation

Strategy automation service

Dynamic resource service

operations and maintenance analysis service

API GW

Business orchestration

Resource orchestration

Network change management

KPI/KQI monitoring …

Design state

AI cloud training

4.2 Intelligent operations & maintenance transformation of a large Southeast Asia telco

• Equipment comes from many different manufacturers, and the operations platform is old. The system has many barriers between different departments, and the operational efficiency is low with high cost.

• The fixed and mobile networks are operated separately and have been unable to be integrated for a long time.

• Customer complaint handling time is longer than expected, and the customer satisfaction rate is low.

• The response to business demands is considered too slow, and TTM cycle is longer than expected.

Operations and Maintenance Challenges

Figure 40: Intelligent operations and maintenance solution architecture of a large telecommunications operator in Southeast Asia

Orchestration services

Algorithms / models are continuously

updated

Event strategies and models are continuously

updated

34Success Stories

Advanced operations and maintenance

scenarios

Basic operations and maintenance

scenarios

old enhancement

One-click diagnosis and recovery of household broadband complaints

Complaint management automation

Fault management automation

Prediction & prevention

OTT dynamic monitoring and diagnosis

Copper wire access prediction and prevention

TT automatically creation

Wireless network health

assessment

TT automatic distribution

WO automatic creation

WO automatic association

WO automatic close

Automatic ChatOps

Optical fiber access prediction and prevention

Wireless hardware prediction and prevention

Analysis of the problem of household broadband users

3G / 4G fault automatic analysis and recovery

plannew

On the other hand, HUAWEI uses the integrated development center of the intelligent operations platform to help the operator organize the application of operations scenarios. As shown in Figure 41, it is mainly divided into four categories:

• Reorganize the existing key operation scenarios in the new operations platform, such as wireless network health assessment, 3G alarm monitoring, 4G alarm monitoring, CS KQI monitoring, terminal analysis, overlap analysis, TT automatic dispatch, TT order automatic association, etc.

• Enhance operational functions, such as traffic analysis, coverage analysis, neighbor analysis, frequency analysis, single user analysis, automatic creation of TT orders, automatic closing of TT orders, etc.

• Build operational applications, such as VoLTE KQI monitoring, VNF alarm monitoring, IMS KPI monitoring, etc.

• Plan operational applications, such as one-key diagnosis and recovery of household broadband complaints, copper-access failure prediction and prevention, fiber-access failure prediction and prevention, wireless-hardware failure prediction and prevention, etc.

The figure below shows the intelligent operations and maintenance application architecture.

Through the reorganization of these operational activities, we helped operators realize skills transformation for operations personnel, hand over repetitive tasks to machines, such as traffic analysis and coverage analysis, and transform the operating model to a new model of human-machine collaboration. Additionally, we have also solved the previous operations problems, such as low operation efficiency and correspondingly high customer complaints. Finally, through organizing Studio, business agility is enhanced and business TTM is shortened.

Figure 41：Intelligent operations and maintenance application architecture of a large telecommunications operator in Southeast Asia

• Improvement of quality: MTTR reduced by 30%–35%. User complaints reduced by 10%. On-site work orders reduced by 5%–20%.

• Reduction of cost: Operations efficiency is expected to increase by 10%–20%, and the number of faults is expected to decrease by 10%–20%.

• Realized Transformation: The operations model has shifted from human-oriented to a new model of human-machine collaboration and has achieved skills transformation for operations personnel.

Benefits to client

Traffic analysis

Coverage analysisNeighborhood

analysis

Frequency analysis

Single user analysis

Critical path analysis

Pilot pollution analysis Service quality

analysis

VAP / VAC / VIP / roaming

analysisTerminal analysis

Overlap analysis

Wireless performance analysis

3G alarm monitoring

VoLTE KQI monitoring

4G alarm monitoring

CS KQI monitoring

VNF alarm monitoring

PS KQI monitoring

FTTx alarm monitoring

CSFB KPI monitoring

IP alarm monitoring

IMS KPI monitoring

SDN alarm monitoring

Service KQI monitoring

NFVi alarm monitoring

VIP KQI monitoring

Power failure alarm

TCP KPI monitoring

35 Summary and Outlook

Summary and Outlook5With the continuous advancement of 5G network deployment and the deep integration of AI in network operations, HUAWEI's human-machine collaborative intelligent operations solution transforms the operations from "human-dependent" to "automated and intelligent." Through skill transformation, the operations personnel are converted into data engineers, network strategists, and application orchestrators. The experts' experience is packaged into operations processes and cognitive assets and injected into the intelligent operations platform. Through the new human-machine collaborative operating model, HUAWEI's intelligent operations solution can break the law of linear resource growth with equipment, use automation to reduce human errors, and improve operations efficiency. Through the adoption of AI technology, HUAWEI's intelligent operations solution enables the prediction and prevention of network and business failures, improves the quality of operations, and contributes to realize the ultra-reliable fourth-generation telecommunications network.

In the future, HUAWEI will continue to use its knowledge and experience in products, technologies, and professional services to bring more "activeness, predictiveness, and preventiveness" into reality and work with the industry to create a healthier and more dynamic telecommunications operations ecosystem. HUAWEI is willing to work with operators to continuously explore and innovate telecommunications network operations, open the HUAWEI's global operations experience, and help operators to build intelligent operations capabilities in order to reduce costs, improve network quality, and enable operations transformation.

36References

References6

• Top 10 Trends Impacting Infrastructure & Operations for 2019 [M]. P10, Gartner, 2019

• Digital Divergence, How European Telcos can cut costs with automation, AI and big data, [M]. Goldman Sachs, 2017


• Industry Expert Interview Notes from 3rd Party

• Research on network optimization based on user NPS [J]. Data communication, issue 5, 2019



• Opportunities and Challenges Coexist, Reducing OPEX Is the Biggest Challenge for Operators [W]. Line 10-12,




• Industrial 4.0 concept and system evolution from 2011 to 2016 [W]. Figure 1,

• Conference of intelligent operation and maintenance technology of Boshi technology 2019 [W]. Line 18-21,

• Conference of intelligent operation and maintenance technology of Boshi technology 2019 [W]. Line 18-21,


• Annual Report of Orange in 2017, [W].

• IG1190 AIOps Service Management Suite – TM Forum

http: //m.elecfans.com/article/701505.html




https: //www.sohu.com/a/123342539_529972

https: //www.passit.cn/hkj/70129.html

https: //www.passit.cn/hkj/70129.html


https: //www.orange.com/en/Investors/Results-and-presentations

https: //projects.tmforum.org/wiki/display/PUB/IG1190+AIOps+Service+Management+Suite+v3.0

General DisclaimerThe information in this document may contain predictive statement including, without limitation, statements regarding the future financial and operating results, future product portfolios, new technologies, 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 © 2020 HUAWEI TECHNOLOGIES CO., LTD. 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.

Tradememark Notice , , are trademarks or registered trademarks of Huawei Technologies Co.,Ltd.Other Trademarks, product, service and company names mentioned are the property of thier respective owners.

HUAWEI TECHNOLOGIES CO., LTD.Huawei Industrial BaseBantian LonggangShenzhen 518129, P. R. ChinaTel: +86-755-28780808www.huawei.com

HUAWEI Intelligent Operations Solution/media/CNBGV2/download/products/serv… · HUAWEI's...

Documents

Transcript of HUAWEI Intelligent Operations Solution/media/CNBGV2/download/products/serv… · HUAWEI's...