A whitepaper of

Autonomous Networks:
Empowering Digital Transformation
For e Telecoms Industry
Objecve(s): To develop a common understanding and consensus on the autonomous
network concept and automaWon classificaWon for the simplificaWon of telecom network
infrastructure, automated & intelligent operaWons and innovaWve services.

Contribu+ng companies and co­editors/contributors:

TM Forum, Aaron Richard Earl Boasman­Patel (Editor)
Huawei Technologies Ltd. Co., Dong Sun (Co­editor)
China Mobile, Ye Wang (Co­editor)
Orange, ChrisWan Maitre (Co­editor)
BT, José Domingos
BT, Yiannis Troullides
Ericsson, Ignacio Mas
Telstra, Gary Traver
Telstra, Guy Lupo

Release 1.0
15 May, 2019

inform.tmforum.org
Contents

1. Objec+ves 3
1.1 Challenges and opportuniWes 3
2. Autonomous networks: framework and levels 5
2.1 Framework 5
2.2 Autonomous networks levels 6
3. Methodology and examples of applica+on scenarios 8
3.1 Methodology and use case mapping 8
3.1.2 Mapping of autonomous network use cases 9
3.2 Examples of applicaWon scenarios 12
3.2.1 O&M Efficiency Improvement 12
3.2.2 Energy Efficiency Improvement 12
3.2.3 Resource Efficiency Improvement 12
3.2.4 User Experience Improvement 12
3.2.5 Honoring Customer Commitment 13
3.2.6 Working with 3rd party providers 13
Addendum 13
Remark 14
References 14

inform.tmforum.org 2
1. Objectives
To provide a common industry understanding of autonomous networks including a harmonized classificaWon system and
supporWng definiWons for network automaWon.

1.1 Challenges and opportunities

Service providers globally are now well underway with their digital transformaWon with 92.5% of those surveyed
for TM Forum’s Digital TransformaWon Tracker saying that they were in the process of their transformaWon and
nearly a quarter (24%) saying that they were “well on the road and reaping significant benefits”. Yet the majority
(44.5%) were just starWng their journey.

But are service providers ready to meet the explosive demand for connecWng objects rather than people?
According to Huawei’s Global Industry Vision (GIV), by 2025 there will be a total of 100 billion connecWons
around the world. Whilst this represents a great opportunity, service providers may find it difficult to meet this
demand for hyper­connecWvity due to the complexity and fragmentaWon of their network architecture,
deployment and integraWon, the inefficiency of their operaWons and a lack of capable knowledge and skills, which
are full of legacy technologies, methodologies and tools. Moreover, the legacy mindset of “build and operate” –
“first, build the network facility and then think about the operaWons”, implies an ignorance and a lack of respect for
the value and importance of operaWons. Indeed, 5G has the potenWal to add to this complexity unless service
providers embrace automaWon and simplify their networks, their operaWons and their management of systems.

Drivers of Automaon

OpEx rao: 50% - 67% 1.9%

Silo, Fragmented
Complexity TCO > Revenue
75B devices by 2025 OpEx Vs CapEx: 3:1 Last 5 years
Simplificaon
Automaon
Infrastructure & Operaons Cost & Growth
Intelligence
Manual, error-prone Collaboraon
Lack of global Isolaon
Efficiency 3 vs 300+ business collaboraon
OTT Vs Telco to maintain
Autonomous
Networks

AutomaWon thus needs to be a central pillar in operators’ digital transformaWon strategies. According to a 2018
report by Analysys Mason, 56% of CSPs globally have li3le or no automaWon in their networks, but according to
their own predicWons, almost 80% expect to have automated 40% or more of their network operaWons, and one­
third will have automated over 80% by 2022.

In December 2018, TM Forum surveyed 65 service providers from 37 unique companies as part of its “AI and its
pivotal role in transforming operaWons” report. One of the quesWons asked was “what is driving automaWon into
operaWons?” As can be seen in the graphic below, 42% of respondents stated that they would like to introduce
new services that require faster and more complicated responses from operaWons (e.g. on­demand services) than
manual processes can provide. A further 28% stated that automaWon is needed to be able to change processes or
seXngs (e.g. provisioning) more o$en and more quickly than manual processes allow.

inform.tmforum.org 3
Drivers for introducing automaon into operaons Autonomous networks have real potenWal and data
from the World Economic Forum indicates that
3% network automaWon could add $9 billion in
operaWng profit from less frequent and shorter
20.5% network outages.
42%
For service providers to really take advantage of the
benefits of network automaWon, especially as part of
their digital transformaWon strategy, they must act
quickly to avoid being out­maneuvered by more
agile DevOps­driven OTT players and new digital
28% players. Autonomous networks need to take
advantage of AI, big data, cloudificaWon and to
provide fully automated, self­healing and self­
6.5%
opWmizing capabiliWes which operate right across
the stack from resource management to operaWons
We are introducing new technology (e.g. virtualizaon) and maintenance customer experience and service
that necessitates operaons at speeds not feasibile for enablement. For service providers to be successful
manual processes
with their AI and automaWon strategies, they need to
We need to change processes or se ngs (e.g. come together with the wider industry ecosystem to
provisioning) more o!en and more quickly than manual build a common understanding of autonomous
processes allow networks. This should include a harmonized
classificaWon system and supporWng definiWons and
We have new process requirements that are more levels of automaWon maturity to enable automaWon
complicated and change more o!en (e.g. partner to happen quicker across the telecoms landscape. It
onboarding for IoT) than manual processes can cope would also need to remove the many pain points to
with allow faster automaWon and innovaWon.
We would like to introduce new services that require
faster and more complicated responses from operaons
(e.g. on-demand services) than manual processes can
provide

Other
TM Forum, 2018

inform.tmforum.org 4
2. Autonomous networks:
framework and levels
2.1 Framework
To build a more valuable telecoms industry and to take advantage of automaWon and AI, a simplified network
architecture and operaWons automaWon are needed for self­configuraWon, self­healing, self­opWmizing and self­
evolving telecom network infrastructures. Operators need to offer zero wait, zero touch, zero trouble services,
which can provide the best­possible user experience, full lifecycle automaWon and maximum uWlizaWon.

Framework for autonomous networks

Personalized Automated Real me Smart Smart Smart

services O&M experience city home manufactory
Business transformaon
✔ On-demand business
Efficiency: Revenue:
✔ Ecosystem business
Agile opera ons Digital enabling

Customer, ecosystem, partner opera ons (Business) Operaons transformaon

Commerce

- Customer centric
Business automaon - Ecosystem oriented
- Collabora on enabled

Network, Customer facing Services opera ons (Network) Operaons transformaon

Produc on

- Full lifecycle automa on

Network automaon - Data/knowledge driven & pla!orm based
- Knowledge as a service

RFS opera ons RFS opera ons (Architectural) Network transformaon

- Extremely simplified architecture
Autonomous domain x Autonomous domain y
- Autonomous domains collabora on
NE NE Cloud Cloud - Enablement of on-demand produc on
NE NE - Support for collabora ve produc on

The key of network transformaWon is to upgrade from fragmented, silo network element interoperability towards a
closed loop of network autonomous domain based on an extremely simplified network architecture, which create
a foundaWon for operaWons automaWon and to enable on­demand producWon by means of autonomous domain
collaboraWon.

The key of network operaWons transformaWon is to upgrade from legacy customized project­centric approach to a
data/knowledge driven pla2orm based on full lifecycle operaWons automaWon. The most important part of this
transformaWon is a mindset change from a “build­and­operate” to a “design with operate”, and the recogniWon of
the value of operaWons knowledge as a service (KaaS). KaaS is about delivering the right knowledge to the right
person in the right context at the right Wme via desktop, laptop or any mobile device. OperaWons automaWon sits
at the core of producWon efficiency and business agility.

The key of business transformaWon is to upgrade from isolated markeWng to a collaboraWve, on demand,
automated business collaboraWon and ecosystem. Business automaWon involves enabling closed loops for
customer/business/ ecosystem operaWons, normally requiring collaboraWon across mulWple service providers
globally.

inform.tmforum.org 5
 Data & knowledge driven intelligent, simplified networks

Knowledge as a service
Design Rollout Provisioning Assurance

Network Full lifecycle Service Full lifecycle

Planning Op miza on Fulfillment Self-configured, self-healing,

Network opera ons as a service self-op mized
Service Cloud

Full lifecycle closed-loop

Cross domain opera ons
Autonomous
Domain Domain Domain
opera ons opera ons opera ons Networks
Autonomous Autonomous Autonomous Levels
domain x domain y domain z

Simplified architecture
E2E SR/SRv6 Best user experience, full lifecycle
vRCBRAS EdgeDC vEPCIMS Regional DC automa on, maximum u liza on
Broadband
2B Backbone/DCI
4G/5G CO Fabric IGW
MAN Fabric
OTN

In order to measure the maturity and capabiliWes of autonomy for the simplified architecture, it is important to
agree on and to properly define what the different categories or levels such as knowledge­as­a­service and full
lifecycle closed loop should be.

2.2 Autonomous networks levels

With the goal of providing a common understanding and consensus for autonomous networks, this whitepaper
delivers a harmonized classificaWon system and supporWng definiWons that:

• Define the concept of autonomous networks

• IdenWfy six levels of network automaWon from “no automaWon” to “full automaWon”.

• Base definiWons and levels on funcWonal aspects of technology.

• Describe categorical disWncWons for a step­wise progression through the levels.

• Educate a wider community by clarifying for each level what role (if any) operators have in performing the
dynamic network operaWons task while a network automaWon system is engaged.

inform.tmforum.org 6
 Autonomous networks levels

L0: Manual L1: Assisted L2: Par3al L3: Condi3onal L4: High L5: Full
Level Defini3on Opera3on & Opera3on & Autonomous Autonomous Autonomous Autonomous
Maintenance Maintenance Network Network Network Network

Execu3on P P/S S S S S

Awareness P P P/S S S S

Analysis P P P P/S S S

Decision P P P P/S S S

Intent/Experience P P P P P/S S

Applicability N/A Select scenarios All scenarios

P: Personnel, S: Systems

Level 0 ­ manual management:

The system delivers assisted monitoring capabiliWes, which means all dynamic tasks have to be executed manually.

Level 1 ­ assisted management:

The system executes a certain repeWWve sub­task based on pre­configured to increase execuWon efficiency.

Level 2 ­ par*al autonomous network:

The system enables closed­loop O&M for certain units based on AI model under certain external environments.

Level 3 ­ condi*onal autonomous network:

Building on L2 capabiliWes, the system with awareness can sense real­Wme environmental changes, and in certain
network domains, opWmize and adjust itself to the external environment to enable intent­based closed­loop
management.

Level 4 ­ high autonomous network:

Building on L3 capabiliWes, the system enables, in a more complicated cross­domain environment, analyse and
make decision based on predicWve or acWve closed­loop management of service and customer experience­driven
networks.

Level 5 ­ full autonomous network:

This level is the ulWmate goal for telecom network evoluWon. The system possesses closed­loop automaWon
capabiliWes across mulWple services, mulWple domains, and the enWre lifecycle, achieving autonomous networks.

The lower levels can be applied now and deliver immediate cost and agility benefits in certain scenarios. An
operators can then evolve to the higher levels, gaining addiWonal benefits and addressing a wider range of
scenarios.

Network automaWon is a long­term objecWve with step­by­step processes, from providing an alternaWve to
repeWWve execuWon acWons, to observing and monitoring the network environment and network device status,
making decisions based on mulWple factors and policies, and providing effecWve percepWon of end­user
experience. The system capability also starts from some service scenarios and covers all service scenarios.

This transformaWon will take several years to fully develop, so we are following an evoluWonary process of
gradually introducing automaWon with AI abiliWes into different domains to bring immediate value.

inform.tmforum.org 7
3. Methodology and examples of
application scenarios
3.1 Methodology and use case mapping
AutomaWng the business should be done incrementally over a period of Wme. It is important to establish operaWng
pa3erns where tradiWonal operaWons can co­exist with automaWon, and where the flow of informaWon does not
have to change as more and more processes become automated. One such approach is to empower the DevOps
teams to:

• idenWfy the network automaWon use case discovery

• create the cost and benefit analysis

• prioriWze their deployment through the natural product backlog processes

• establish the discipline of determining the informaWon flow of data which informs both automaWon and non­
automated processes to be the same

The following two aspects are criWcal for implemenWng the network automaWon design.

a. The need for an arbitraWon policy in achieving network automaWon. As processes are automated, the decision
rights on how decisions are administered should be data driven and must be applied through policy. The
following rules must be well understood and transparent: a) who has decision rights on the policy? B) what are
the condiWons and resulWng acWons that must be taken? And, c) where is the policy is applied?

b. The need for an e2e product­service­resource inventory topology view, and its potenWal pa3erns. This includes
the ability to apply resources when needed whether they are for new business, service restoraWon, or for
healing broken segments. It requires the ability to quickly gain an understanding of the service inventory
available, and of whatever prioriWzaWon rules that must be in place to apply on these resources.

The selecWon of use cases for autonomous networks can be based on a combinaWon of technical feasibility,
business orientaWon, capability promoWon, and value priority, as shown in the following figure.

Funnel of analysis for autonomous networks use cases

Business Orientaon Principles Technical Feasibility Principles

1. Whether need to contact user directly? 1. Is the technology mature?
2. Is there a high failure rate and Network O&M 2. Is there any major risk in the
complaint rate? process technology?
3. Is there a large amount of manpower or 3. Does the technology have similar
energy cost? Mature technology success case?
4. Whether need much me? process/acvies 4. Can the technology reach the manual
level?
High service value
Capability Promoon Principles process/acvies Value priority principle
1. Is there a similar requirement? Promoon in 1. Priority 1: Technical is feasible, service
2. Is there a similar basic service process? cross-province value is high, and convenient promoon
process 2. Priority 2: The technology is feasible
3. Is there a similar data structure? acvies and the business value is high
4. Whether the technology is mastered Comprehensive
independently? 3. Priority 3: The technology is feasible
value ranking

inform.tmforum.org 8
In terms of the evoluWon and deployment of autonomous network use cases, each company can contribute use
cases in the context of key funcWons in the eTOM process framework as the examples.

Mapping of autonomous network use cases

Strategy, infrastructure & product Opera ons

Strategy & Infrastructure Product lifecycle Opera ons Fulfillment Assurance Billing & revenue
Commit lifecycle management support & management
management readiness
5G Core 5G slicing Root cause
...... network ...... ...... analysis for ......
configura on
planning network faults

...... ...... Network

traffic forecast

......

Enterprise management

3.1.2 Mapping of autonomous network use cases

Using the eTOM process framework and an automaWon classificaWon of autonomous networks, to create a funnel
analysis and AI technology, we have developed a map of autonomous network use cases as shown in the
following table.

Mapping of autonomous network use cases

L0: Manual L1: Assisted L2: Par3al L3: Condi3onal L4: High L5: Full
Level Defini3on Opera3on & Opera3on & Autonomous Autonomous Autonomous Autonomous
Maintenance Maintenance Network Network Network Network

Complete External Experience

Automate Complete awareness, environments awareness to
O&M person repeated acWons network analysis, awareness to closed­loop
manually or redundant element closed­ decision­making closed­loop O&M and
Typical Features
executes all informaWon loop control and execuWon O&M across learning for
dynamic tasks. eliminaWon base on AI closed­loop domains with cross­industry
based on rules. models. base on AI self­learning public ICT
system. capabiliWes. infrastructure.

Customer
Per To­B\To­C business journey and experience measurement metrics
experience journey

Opera3ons Based on network planning, construcWon, network operaWons & maintenance, opWmizaWon, and business operaWon
processes acWviWes, through AI model training and system opWmizaWon to implement process integraWon and automaWon

Network
Converged Wme
management
and space
control , and
service
Telecom AI analysis
training pla2orm capability
SeparaWon of
Networks Public ICT
ExisWng networks and devices services and
Capabili3es Hierarchical AI Integrated & Infrastructure
networks
inference wireless
module backhaul
Network device­
and cloud
Time and space
CollaboraWon
capability

inform.tmforum.org 9
 L0: Manual L1: Assisted L2: Par3al L3: Condi3onal L4: High L5: Full
Level Defini3on Opera3on & Opera3on & Autonomous Autonomous Autonomous Autonomous
Maintenance Maintenance Network Network Network Network
Experience
driven wireless
opWmizaWon
Intelligent
Experience
idenWficaWon of Personal Digital
Experience Wireless driven network
encrypted Twins
network opWmizaWon
services
parameter
opWmizaWon

Base StaWon
beam forming
Radio resources Network Resource
Virtual radio
Resource needed on resource needed needed on
grids Network traffic
demand. on demand demand
forecast

Public ICT
DC PUE
Dynamic infrastructure
(Water­cooled) Dynamic DC
Use Periodic wireless power saving
power saving
Cases Energy shutdown of the network power per dynamics of
Intelligent per
BTS saving per traffic the usage,
shutdown of the environments
usage environments
BTS
etc.

One trouble one

Wcket
Self­recovery of
Radio
Manual upgrade Transmission ower faults
neighboring cell
opWcal layer
configuraWon Automated
Manual Cutover commissioning Feeder fault
cross­Domain
eliminaWon
O&M WiFi parameter Service
Manual recovery Automated
opWmizaWon Provisioning
of BTS power recovery of BTS Fiber fault
(VPN)
failure power failure predicWon and
Core Network
PON OpWcal locaWng
changing
Link trouble
PredicWon and
locaWon

inform.tmforum.org 10
 L0: Manual L1: Assisted L2: Par3al L3: Condi3onal
L4: High Autonomous L5: Full Autonomous
Level Defini3on Opera3on & Opera3on & Autonomous Autonomous
Network Network
Maintenance Maintenance Network Network

Services take
days to Services take hours
Services take
provision. to provision.
Services take minutes to
days to provision.
Services are Services and
provision.
Services take exposed via an changes to those
Intent is provided
weeks to online service services are fully
Services are by the customer
provision catalogue with expressed in intent
Services take ordered online interacWng
basic modelling by the customer
months to from a through Service
Services are capabiliWes of through a Service
provision catalogue catalogue APIs
online ordered services. catalogue
and infrastructure
but fulfillment
Services are Some level of is fully intent
process remains Customer can Services are
manually Intent is used driven.
largely manual interact with assigned and
defined for strategic
catalogue and designed
so$ware Auto­generated
Parts of the order service automaWcally from
Intent Then ordered defined services set of workflow
fulfillment are components intent
Driven by customers steps based on
automated specificaWons in
Services Some Services the intent
through staWc Services consist the catalogue
and then are acWvated expressed in the
hard coded of mulWple
manually automaWcally catalogue.
workflows service Resources are
designed
components and dynamically created
Customer can Service is
Only a few end resources on demand.
and then change many automaWcally
service changes
manually service provisioned and
are possible Parts of the Intent is translated
provisioned parameters assured.
online with service are into automaWc
The Autonomous Network Capabili3es Use Cases

online via the

mostly manual automaWcally network acWvaWon,
catalogue and Network is
provisioning configured but some aspects
these are primarily
remain staWc with
automaWcally composed of
Customer can some manual work
provisioned and virtualized
change some of required (for
assured infrastructure
the service network physical
parameters infrastructure).
online.

Services take Services never fail

Services take
seconds to restore
minutes to
Self­opWmised,
Services take restore
Automated closed Self­healing
hours to fix
loop end2end network
AI and Machine
Automated learning driven
Large parts of the Fully zero touch
Services take orchestraWon automaWon
autonomous management
days to fix
network are fully
Many of the Real­Wme
automated with The network is
Centralised network telemetry
Services take closed loop control fully Autonomous
control of some features are working with AI
weeks to fix happening based and control
of the parts of centrally systems
on real­Wme views systems work in
the network controlled with providing offline
Network is of telemetry closed loop
Telemetry allowing some available analyWcs
largely permanently, with
and control interfaces. capabiliWes
composed of Fully real­Wme views of
Closed­ capabiliWes
distributed programmable the network
loop Some repair AI models with
systems and interfaces allow available.
Control Network is funcWons are analyWcs
has no so$ware control of
primarily automaWcally provide a
centralized the autonomous InformaWon is
manually performed with framework for
control network. DevOps analysed in real­
repaired, but zero human closed loop
fully implemented. Wme by AI expert
some repair intervenWon. control with
systems and
funcWons can be some network
Real­Wme insights changes happen
remotely Closed loop restore
on network in split seconds
performed control becomes processes being
performance are automaWcally.
possible with possible.
automaWcally
some real Wme
created and No manual
telemetry DevOps control
provided to intervenWon is
available of some
operaWons – required except in
network aspects
DevOps. extreme
circumstances.

inform.tmforum.org 11
3.2 Examples of application scenarios
In the previous secWon, we outlined the methodology and mapping of use cases selecWon. Based on these criteria,
we have selected the following scenarios for the purpose of illustraWon and clarificaWon.

3.2.1 O&M Efficiency Improvement

In the O&M domain, O&M pracWces can be divided into three categories based on how proacWve or passive they
are. The first category is called Run­to­Failure (R2F). If there is a fault on a network, the O&M personnel
immediately go to the site and recWfy the fault. This is the lowest level. The second category is called PrevenWve
Maintenance (PvM). With PvM, each device is checked to prevent faults, but efficiency is low. The third category
is called Predictable Maintenance (PdM). With PdM, we can calculate the probability that a device will become
faulty in the future and then perform targeted maintenance. With PdM, we hope to reduce the workload required
for alarm handling and fault locaWng on telecoms networks by 90%. We also aim to predict the failures and
deterioraWon of 90% of key components, and to achieve network self­healing. In addiWon, more than 70% of the
problems in network faults lie in passive equipment, such as aging or bent opWcal fibers, and ports that have
become loose. In all of these situaWons, signals change. AI is introduced to learn the characterisWcs of these
changes, to enable operators to predict the changes in advance and to use soluWons to solve passive faults. Fault
locaWon and predicWon for PON opWcal modules and opWcal links provides an instrucWve example of the effect AI
can have on O&M pracWces.

3.2.2 Energy Efficiency Improvement

When it comes to energy efficiency, the number of bits should determine the number of wa3s. That is, lower
traffic volume should result in lower power consumpWon. In an equipment room or at a site, each system is
configured with dozens of parameters. The heat dissipaWon, environment, and service load models are generated
through AI training to maximize the energy consumpWon efficiency for sunlight, temperature, and auxiliary
faciliWes, such as diesel generators, solar energy devices, and ba3eries. At the equipment layer, dynamic energy
distribuWon is performed based on service loads. If there is no traffic, power consumpWon is reduced by using
Wmeslot shutdown, RF deep sleep, and carrier shutdown. In addiWon, energy consumpWon of data center
hardware, such as server components, can be dynamically reduced. The power consumpWon of the network
system must also be taken into account. The accurate service load predicWon model is constructed to balance the
traffic on the enWre network and achieve the opWmal energy consumpWon efficiency.

3.2.3 Resource Efficiency Improvement

Regarding network resources, on most present­day networks the flow of traffic simply follows the physical
topology of the network, and resource usage may be unreasonable. If network scheduling took into account the
direcWon of traffic flow, resource usage efficiency could be greatly improved. At present, however, networks do
not have this capability. To address this issue, AI must be introduced, and a traffic predicWon model must be
created. In this way, precise traffic predicWon and the opWmal network topology can be provided, and the network
paths can be determined by traffic instead of just by physical connecWons.

3.2.4 User Experience Improvement

Network design, planning, and configuraWon are automaWcally performed and the service rollout Wme can be
shortened to one tenth of the original Wme. For individual users, device and applicaWon data is used to achieve
opWmal user experience based on the automated closed­loop mechanism. For enterprise users, mulW­domain
global networks can be deployed in minutes with cloud­network synergy enabled, and network­wide routes are
automaWcally adverWsed and learned, implemenWng scheduling and rouWng in ways that are invisible to users. For
home users, device­cloud synergy improves home broadband quality and extends experience improvement and
problem handling to home networks.

inform.tmforum.org 12
3.2.5 Honoring Customer Commitment
Using the service data to ascertain whether or not we are meeWng our customer commitment (SLA), maintain
Service Level ObjecWves (SLO) proacWvely. From this and other data can discern the level of saWsfacWon the
customer has with the service and what steps we can take to enhance the relaWonship with the customer.

3.2.6 Working with 3rd party providers

As services are provisioned to customers, some of the components of those products or services are supplied by
third parWes. UWlizing the same abstracWon framework as deployed with Domains, each third party provider can
have their elements of the service integrated and automated into the overall service mix. Securing agreement to
share telemetry enhances the ability to perform a consolidated closed loop assurance soluWon improving our
customer responsiveness.

Addendum:
Telstra’s Network­as­a­service case study
Telstra has gained valuable experience of network automaWon through its network­as­a­service project. Full
details of the project can be found at h3ps://inform.tmforum.org/casestudy/bringing­telstra­naas­life/

To simplify the network Telstra used the approach of carving up the network into a number of different “Network
Service Domains” Each domain has a defined set of capabiliWes it performs. Each Domain must accept a set of
responsibiliWes which enable it to have services “consumed” in an automated way through a standard service
catalog. What sits in front of the domain is NaaS (Network as a Service) an abstracWon layer which enables any
approved enWty to request services from those domains using a standardized framework common to all domains.

The network as a Service “NaaS” APIs are important ingredients to design simplified network, closed loop
operaWon and business automaWon. The NaaS abstracWon layer exposes TMF Open APIs for service instance
management operaWons. It facilitates communicaWon between:

1. Network service domains and Digital Apps, BSS/OSS

2. Across mulWple Network Service Domains

NaaS allows Digital & IT systems to communicate with network services domains using standardised language and
constructs for managing the network services instance lifecycle.

It also facilitates cross network domains interacWons for execuWng various operaWons including feasibility,
provisioning, acWvaWon, problem noWficaWon, diagnosWcs and remediaWon. It also provides a communicaWon
mechanism between two or more network service domains for managing composite service instance operaWons.

inform.tmforum.org 13
Remark:
1. The white paper will be refreshed periodically in the form of new releases per the contribuWons.

2. In parWcular, the secWon of use cases is evolved per further contribuWons and refinements, for instance, the use
cases may be aligned based on the closed loop of network automaWon, operaWons automaWon and business
automaWon.

References:
1. AI maturity model

inform.tmforum.org 14