ETSI GR ENI 007 V1.1.1 (2019-11)

Experiential Networked Intelligence (ENI); ENI Definition of Categories for AI Application to Networks

DGR/ENI-0011_Def_Cat_AI_Ap_Net

General Information

• Status: Published
• Publication Date: 17-Nov-2019
• Technical Committee: ENI - Experiential Networked Intelligence
• Current Stage: 12 - Completion
• Due Date: 28-Nov-2019
• Completion Date: 18-Nov-2019

Standards Content (Sample)

GROUP REPORT
Experiential Networked Intelligence (ENI);
ENI Definition of Categories for AI Application to Networks
Disclaimer
The present document has been produced and approved by the Experiential Networked Intelligence (ENI) ETSI Industry
Specification Group (ISG) and represents the views of those members who participated in this ISG.
It does not necessarily represent the views of the entire ETSI membership.

2 ETSI GR ENI 007 V1.1.1 (2019-11)

Reference
DGR/ENI-0011
Keywords
artificial intelligence, categorization, category,
network
ETSI
650 Route des Lucioles
F-06921 Sophia Antipolis Cedex - FRANCE

Tel.: [phone]  Fax: [phone]

Siret N° 348 623 562 00017 - NAF 742 C
Association à but non lucratif enregistrée à la
Sous-Préfecture de Grasse (06) N° 7803/88

Important notice
The present document can be downloaded from:
http://www.etsi.org/standards-search
The present document may be made available in electronic versions and/or in print. The content of any electronic and/or
print versions of the present document shall not be modified without the prior written authorization of ETSI. In case of any
existing or perceived difference in contents between such versions and/or in print, the prevailing version of an ETSI
deliverable is the one made publicly available in PDF format at www.etsi.org/deliver.
Users of the present document should be aware that the document may be subject to revision or change of status.
Information on the current status of this and other ETSI documents is available at
https://portal.etsi.org/TB/ETSIDeliverableStatus.aspx
If you find errors in the present document, please send your comment to one of the following services:
https://portal.etsi.org/People/CommiteeSupportStaff.aspx
Copyright Notification
No part may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying
and microfilm except as authorized by written permission of ETSI.
The content of the PDF version shall not be modified without the written authorization of ETSI.
The copyright and the foregoing restriction extend to reproduction in all media.

© ETSI 2019.
All rights reserved.
DECT™, PLUGTESTS™, UMTS™ and the ETSI logo are trademarks of ETSI registered for the benefit of its Members.

3GPP™ and LTE™ are trademarks of ETSI registered for the benefit of its Members and
of the 3GPP Organizational Partners.
oneM2M™ logo is a trademark of ETSI registered for the benefit of its Members and
of the oneM2M Partners. ®
GSM and the GSM logo are trademarks registered and owned by the GSM Association.
ETSI
3 ETSI GR ENI 007 V1.1.1 (2019-11)
Contents
Intellectual Property Rights . 4
Foreword . 4
Modal verbs terminology . 4
1 Scope . 5
2 References . 5
2.1 Normative references . 5
2.2 Informative references . 5
3 Definition of terms, symbols and abbreviations . 6
3.1 Terms . 6
3.2 Symbols . 6
3.3 Abbreviations . 6
4 Overview . 6
4.1 Background on AI integration and network autonomicity . 6
4.2 Examples of application in specific areas . 8
4.2.1 Automation applied to vehicle driving. 8
4.2.2 Autonomicity applied to home broadband services . 9
5 Categories of Network Autonomicity . 10
5.1 Introduction . 10
5.2 Factors determining the network autonomicity level . 11
5.2.1 Technical factors . 11
5.2.2 Market factors . 12
5.3 Network autonomicity categories description . 12
5.4 Tabular representation of Network Autonomicity categories . 13
5.5 Scenario Examples of Network Autonomicity categories . 16
5.5.1 Introduction. 16
5.5.2 Autonomicity categories of network traffic classification . 16
5.5.3 Autonomicity categories of IDC energy management . 19
5.5.4 Autonomicity categories for network fault recovery . 21
5.5.5 Autonomicity categories of DCN service and resource design . 24
5.5.6 Autonomicity categories of DCN service quality optimization . 27
5.5.7 Autonomicity categories of End-to-End service quality assurance in bearer network . 30
6 Relation of the Autonomicity Categories to ENI system architecture and other architectures . 33
6.1 Introduction . 33
6.2 Mapping of Assisted System Classes into Network Autonomicity Categories . 33
7 Conclusions . 35
Annex A: Related work published by other SDOs . 36
History . 37

ETSI
4 ETSI GR ENI 007 V1.1.1 (2019-11)
Intellectual Property Rights
Essential patents
IPRs essential or potentially essential to normative deliverables may have been declared to ETSI. The information
pertaining to these essential IPRs, if any, is publicly available for ETSI members and non-members, and can be found
in ETSI SR 000 314: "Intellectual Property Rights (IPRs); Essential, or potentially Essential, IPRs notified to ETSI in
respect of ETSI standards", which is available from the ETSI Secretariat. Latest updates are available on the ETSI Web
server (https://ipr.etsi.org/).
Pursuant to the ETSI IPR Policy, no investigation, including IPR searches, has been carried out by ETSI. No guarantee
can be given as to the existence of other IPRs not referenced in ETSI SR 000 314 (or the updates on the ETSI Web
server) which are, or may be, or may become, essential to the present document.
Trademarks
The present document may include trademarks and/or tradenames which are asserted and/or registered by their owners.
ETSI claims no ownership of these except for any which are indicated as being the property of ETSI, and conveys no
right to use or reproduce any trademark and/or tradename. Mention of those trademarks in the present document does
not constitute an endorsement by ETSI of products, services or organizations associated with those trademarks.
Foreword
This Group Report (GR) has been produced by ETSI Industry Specification Group (ISG) Experiential Networked
Intelligence (ENI).
Modal verbs terminology
In the present document "should", "should not", "may", "need not", "will", "will not", "can" and "cannot" are to be
interpreted as described in clause 3.2 of the ETSI Drafting Rules (Verbal forms for the expression of provisions).
"must" and "must not" are NOT allowed in ETSI deliverables except when used in direct citation.

ETSI
5 ETSI GR ENI 007 V1.1.1 (2019-11)
1 Scope
The present document defines various categories for the level of application of Artificial Intelligence (AI) techniques to
the management of the network, going from basic limited aspects, to the full use of AI techniques for performing
network management.
2 References
2.1 Normative references
Normative references are not applicable in the present document.
2.2 Informative references
References are either specific (identified by date of publication and/or edition number or version number) or
non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the
referenced document (including any amendments) applies.
NOTE: While any hyperlinks included in this clause were valid at the time of publication, ETSI cannot guarantee
their long term validity.
The following referenced documents are not necessary for the application of the present document but they assist the
user with regard to a particular subject area.
[i.1] ETSI GS NFV 003 (V1.3.1): "Network Functions Virtualisation (NFV); Terminology for Main
Concepts in NFV".
[i.2] MEF PDO CfC (V0.8): "Policy-Driven Orchestration", September 2019.
[i.3] ETSI GS ENI 001 (V2.1.1): "Experiential Networked Intelligence (ENI); ENI use cases".
[i.4] MEF 55: "Lifecycle Service Orchestration (LSO): Reference Architecture and Framework",
March 2016.
[i.5] MEF MCM 78: "MEF Core Model", September 2019.
[i.6] Gamma E., Helm R., Johnson R. and Vlissides J.: "Design Patterns: Elements of Reusable Object-
Oriented Software", Addison-Wesley, November 1994. ISBN 978-0201633610.
[i.7] ISO/IEC 2382-28: "Information technology -- Vocabulary".
[i.8] ISO/IEC/IEEE 42010: "Systems and software engineering -- Architecture description".
[i.9] ETSI GR ENI 004: "Experiential Networked Intelligence (ENI); Terminology for Main Concepts
in ENI".
[i.10] ETSI GS ENI 005 (V1.1.1): "Experiential Networked Intelligence (ENI); System Architecture".
[i.11] ETSI GS ENI 002 (V2.1.1): "Experiential Networked Intelligence (ENI); ENI requirements",
September 2019.
[i.12] ETSI GR ENI 003 (V1.1.1): "Experiential Networked Intelligence (ENI); Context-Aware Policy
Management Gap Analysis", May 2018.
[i.13] TM Forum whitepaper of Autonomous Networks: "Empowering Digital Transformation For The
Telecoms Industry".
NOTE: Available at https://www.tmforum.org/wp-content/uploads/2019/05/22553-Autonomous-Networks-
whitepaper.pdf.
ETSI
6 ETSI GR ENI 007 V1.1.1 (2019-11)
[i.14] 5G-PPP White Paper: "5G Automotive Vision", October 20, 2015.
SAE document J3016: 'Taxonomy and Definitions for Terms Related to On-Road Automated
Vehicles", January 16, 2014.
3 Definition of terms, symbols and abbreviations
3.1 Terms
For the purposes of the present document, the terms given in ETSI GR ENI 004 [i.9] apply.
3.2 Symbols
Void.
3.3 Abbreviations
For the purposes of the present document, the abbreviations given in ETSI GR ENI 004 [i.9] apply.
4 Overview
4.1 Background on AI integration and network autonomicity
ETSI ISG ENI defines how Artificial Intelligence (AI) can be usefully applied in telecommunication networks to
support the management objectives of operators. These include making management faster, more efficient and
providing higher resilience and reliability of the infrastructure and of the services delivered to end-users.
AI can make Operation and Maintenance (O&M) of a traditional network much more efficient with significant cost
savings. AI application for early fault discovery and location, for instance, can enhance the performance of the network
as perceived by end-users as well as by the operator, and reduce fault detection and recovery costs for the operator; this
will in turn reduce loss of income due to service unavailability and from the reduction of maintenance costs thanks to
early fault discovery and location.
The transition to virtual networks will further enhance the benefits of AI application. AI can support network entities as
orchestrators and provide different ranges of management, from assisting and recommending changes (but not actually
performing changes), to performing only those changes that are trusted by the operator, to performing changes without
human intervention. AI can enable the dynamic adaptation of resources to changing traffic conditions and business
goals, and even enable trusted changes without human intervention; this produces a fully self-managed network in
normal conditions. If extraordinary conditions (e.g. when the network exhibits complex faults or is under attack),
external (manual) intervention is required (though the AI can provide recommendations for fixing problems).
Figure 1 illustrates the expected step-by-step evolution of networks as AI is integrated into them as well as trusted by
operators.
ETSI
7 ETSI GR ENI 007 V1.1.1 (2019-11)

Figure 1: Scenarios of network evolution with the integration of AI
Figure 1 shows how the gradual introduction of increased AI functionality enables more management and control
decisions to be enhanced and made with increasingly less human intervention. Most importantly, the use of AI enables
many other functionalities to operate with increased accuracy and effectiveness.
• The traditional implementation of network control and management, with no AI and essentially no automation
is not shown in the figure.
• Manual control requires complete human intervention. Even if AI is present, it is up to the human to make all
decisions and issue all commands. This is sometimes called reactive processing (see clause 4.5.4.2 of ETSI
GS ENI 005 [i.10]).
• The first step of using AI has two notable effects. First, it enhances existing telemetry by interacting with the
network and ensuring that the most applicable telemetry is provided for a given context. Second, it enables the
human to trust its recommendations, and gradually take over some of the management duties. Both of these are
based on the ability of the AI to understand cause-and-effect relationships between monitored data and issued
commands, and more importantly, for the AI to explain its reasoning so that humans trust its conclusions. This
is also called deterministic management (see clause 4.5.4.3 of ETSI GS ENI 005 [i.10]).
• The next step of using AI builds on the previous level, and provides tighter integration with analytics. This
enables the system to move from deterministic to predictive management. Predictive management uses
different processes to calculate probable future events, states, and/or behaviours. Predictive processing also
typically allows probability and/or risk assessment. The prediction is based on analysing current and historical
operation, and applying patterns found to identify possible problems as well as opportunities for improving
operation (see clause 4.5.4.4 of ETSI GS ENI 005 [i.10]).
• The next step of using AI builds on the previous level, and adds elementary goal-directed behavior. Predictions
are now made that are directly related to business goals, which enables the network to provide optimized end-
user services. For example, as user needs, business goals, or network conditions change, the system can
dynamically change the network configuration to adapt to these changes.
• The final step of using AI builds on the previous level, and creates a cognitive management system. While the
previous level is able to make decisions based on stated goals, a cognitive system is able to create new goals in
order to optimize system operation. Cognitive processing enables the system to understand what has happened
and plan a corrective set of actions to achieve system goals and optimize operations (see clause 4.5.4.5 of
ETSI GS ENI 005 [i.10]).
ETSI
8 ETSI GR ENI 007 V1.1.1 (2019-11)
It is recommended that the system uses policy-based management to issue commands, regardless of the level of AI
being used. This is because it provides consistent and auditable behavior. This is critical for enabling adaptive and
flexible service offerings that respond to changing business goals, user needs, and environmental constraints.
It is therefore possible to speak of different categories, of implementation of AI in telecommunications networks,
according to how much the AI system will be able to influence the adaptation of the network and to what extent the
diverse parts of the network are controlled using AI assistance.
4.2 Examples of application in specific areas
4.2.1 Automation applied to vehicle driving
The concept of automation categories is already used in specific fields, such as the automotive industry [i.14], for which
the aspects considered to define the categories involve the level of intervention required by the human driver of a
vehicle, versus the level of control delegated to networked, onboard, and environmental AI assisted agents. The
categories of vehicle driving automation have been defined by the Society of Automotive Engineers (SAE), USA and
by the German Association of the Automotive Industry (Verband der Automobilindustrie (VDA)). The levels of
automation defined by the SAE/VDA for automotive applications are illustrated in Figure 2, where the description of
the level of control the driver has over the car for each category is also given.

NOTE: Reproduced with the permission of the 5G-PPP www.5g-ppp.eu.

Figure 2: Levels of automation defined by the SAE/VDA for automotive applications [i.14]

NOTE 1: The content of the figure refers to the terminology used by SAE/VDA in the context of automotive
applications, e.g. the term automation may often be found. Therefore, Figure 2 is provided just as a
conceptual example and the used terminology is not relevant to the present document.
ETSI
9 ETSI GR ENI 007 V1.1.1 (2019-11)
Analysing the categories defined in the figure, and the associated documents [i.1] and [i.2], one can easily understand
how categories can also be defined for an autonomic network, if the role of the car driver is replaced by that of the
human operator of a network. With reference to Figure 2, instead of 'driver in the loop', one could speak of 'operator in
the loop'. Moreover, increasing categories will have an increasing role of autonomic management based on AI
application, with the subsequent reduction of required intervention of the operator. ETSI ISG ENI already provides the
fundamentals for defining different categories of AI based network autonomicity in the "Use Cases" [i.3],
"Requirements" [i.11] and in the "Context-aware policy management Gap Analysis" [i.12] documents. It is therefore
straightforward to define AI-based network-autonomicity categories, in analogy to what has been just described in the
different area of automotive application for AI. However, as it has been pointed out in note 1, the car automation
example is not strictly and directly mapped to any network autonomicity use cases.
NOTE 2: An autonomic system is a self-governing system that can manage itself in accordance with high-level
objectives.
NOTE 3: Self-governance is performed using cognition. The term "manage itself" means that the autonomic system
can sense changes in itself and its environment, analyse changes to ensure that business goals are being
met, and execute changes to protect business goals.
4.2.2 Autonomicity applied to home broadband services
The example in this clause is the application of the concepts discussed above to a specific and limited set of network
resources; this is the home network scenario, which is similar to an operator's network on a much smaller scale and with
a much simpler architecture.
Figure 3 illustrates the definition of categories for autonomicity of broadband home services.
Figure 3: Example of definition of categories for autonomicity of home broadband services
NOTE 1: Figure 3 was originally made for different purposes, and is provided in the present document as an
example; the terms used in it do not match necessarily the terminology used in ENI.
• The L0 category, fully manual, does not appear in figure 3.
• In the L1 category, where the Network Management System (NMS) is used to deliver device configuration
scripts in batches, the efficiency is improved.
• In the L2 category, the Optical Network Terminal (ONT) box supports plug and play (PnP) for home terminals
and implements partial autonomicity of the home network and services management. An example of this
category is the implementation of intelligent fault location for the access network, at the level of the Optical
Distribution Network (ODN).
NOTE 2: In addition, zero touch deployment, where devices are provisioned and configured automatically, also
avoids the use of manual operation.
ETSI
10 ETSI GR ENI 007 V1.1.1 (2019-11)
• In L3 category, conditional autonomicity, the system is autonomic in some phases of the service life cycle. For
example, the service provisioning, driven by the intent 'broadband service', could configure 100 Mbit/s of
fixed bandwidth based on the type of the house (villa, apartment), or could even define the bandwidth to be
allocated based on user preferences (such as degree of game enthusiasm and video enthusiasm). This
implementation completely replaces manual service deployment and greatly improves the O&M efficiency in
service deployment.
NOTE 3: Home Wi-Fi self-healing and fault self-location and dispatch may also be utilized.
• The L4 category is highly autonomic, where service awareness and decision-making based on service
provisioning are implemented. For example, with this increase on the degree of autonomicity, proactive
maintenance, SLA-based self-healing, and service driven self-organizing networks can be implemented. This
could represent a solution in support of those scenarios for which continuously accelerating service innovation
creates revenue, while planning and design are not required.
• In the L5 category, full autonomic management is implemented in all scenarios, which can be seen as a long
term goal.
The concepts provided in the examples will have to be extended in order to define general categories for the AI based
networks autonomicity. This will be done in the following clauses.
5 Categories of Network Autonomicity
5.1 Introduction
Establishing suitable network autonomicity categories is helpful to guide users in choosing a specific implementation of
AI assisted network, and understanding the self-adaptation capabilities to, i.e. changed service conditions, faults,
deployment of new services and the autonomicity of operation and overall management.
It is important to note that the categories are focusing on the level of autonomicity the network is characterized by as a
consequence of AI tools deployment. In other words, the different categories are providing a classification in terms of
the advantages automation and autonomicity bring into the network management and operation processes thanks to the
capabilities of adaptation and optimization acquired. Different categories will therefore correspond to different
approaches and perspectives in network management and operation. The lowest category corresponds to the absence of
automation and autonomicity, and their presence and influence increase in higher categories up to the full AI based
autonomicity of network management and operation, thanks to cognition capabilities and closed loop implementation.
In the present clause, the characteristics that an AI assisted network needs are discussed with respect to:
1) match the requirements imposed by the services that the provider wants to implement; and
2) offer the needed level of autonomicity (which includes the many aspects mentioned above in clause 4).
Categories of network autonomicity are detailed below with a clear indication of what services and applications and
what management approach the network is suited to support.
ETSI
11 ETSI GR ENI 007 V1.1.1 (2019-11)
5.2 Factors determining the network autonomicity level
5.2.1 Technical factors
A number of technical factors need to be taken into account to determine the degree of AI based autonomicity in a
network, and thus, the category that the system can be associated with. The increasing autonomicity of the network with
the support of the ENI system open different market perspectives for its use in terms of supported services. Therefore,
both technical and market factors will be taken into account for the different categories of network autonomicity,
offering separate views. In this clause the technical factors will be considered, while market factors are considered in
clause 5.2.2. The technical factors that influence the AI assisted network autonomicity are:
• Man-machine interface and level of required manual interaction/configuration with:
- Imperative based mode, requiring to specify how a change in the network configuration is implemented
(e.g. all manual, CLI by-device configuration, NETCONF multi-device collaboration management)
- Declarative/intent based mode, requiring to specify what should be configured, without specifying how
this configuration is realized
• Data collection and awareness parameters that define the attainable level of awareness:
- Single device and shallow awareness (e.g. based on SNMP events and alarms)
- Local awareness (e.g. based on SNMP events, alarms, KPIs, and logs)
- Comprehensive awareness (based on basic telemetry data)
- Comprehensive and adaptive sensing (such as compatible with data compression and optimization
technologies)
- Adaptive posture awareness (edge collection plus judgment)
- Adaptive optimization upon deterioration (edge closed-loop, including collection, judgment, and
optimization)
• Decision making participation (human operator in the loop or not): relevance of human decision versus
machine decision
• Degree of intelligence and level of knowledge, analysis and understanding:
- Lack of understanding (manual analysis)
- Limited autonomous analysis
- Deep autonomous analysis
- Comprehensive knowledge based short-term forecast
- Comprehensive knowledge based long-term forecast
- Self-adaptation and knowledge-based reasoning
• Adaptation of the configuration to changes in the environment:
- Static if no autonomic adaptation is supported
- Limited adaptability to changes
- Adaptability to significant changes
- Any change when any autonomic adaptation is supported
ETSI
12 ETSI GR ENI 007 V1.1.1 (2019-11)
• Supported scenarios including complexity and type of domain, use cases and architecture:
- Single scenario
- Selected scenarios
- Multiple scenarios
- Any scenario
5.2.2 Market factors
The factors that impact the market relevance of network autonomicity involve the possibility to adapt the system and
create service offers in different scenarios and involving, according to the 5G network concept, different stakeholders
covering a part of or the whole service chain. The market relevance is determined by aspects as the level of simplicity
of the AI assisted Network management, the resulting flexibility of the supported services, the required effort and
staffing to operate and manage the network, the usage of resources and energy, the level of customer experience.
There are several factors that impact the simplicity of management and control of the AI assisted network and therefore
the market relevance that an AI system (enabling the given level of simplicity) can have. An example can be the service
creation and configuration interface, which is based on basic command line and complex procedures or an intent based
automated one; another is that of a Network Management System (NMS) that, when combined with a controller, is able
to realize the full automation in a single domain and/or cross-domain. Examples of such factors include (but are not
limited to):
• Scheduling execution: implemented by a person (i.e. manually executed), person and system
(i.e. semi-automated), or system (i.e. fully automated).
• Network perceptual monitoring: performed by a person, person and system, or system.
• Analysis and decision-making: done by a person, person and system, or system.
• Service recovery: operated by a person, person and system, or system.
• Customer experience evaluation: by a person, person and system, or system.
• Service coverage capability: coverage of some service scenarios or all service scenarios.
It is challenging to provide a comprehensive list of all the relevant factors. Some of the most relevant are used in
Table 2 of clause 5.4 to provide an example of the application of the concept, which is not considered exhaustive.
5.3 Network autonomicity categories description
From totally operator controlled up to the full autonomicity of network management and control based on AI, the
categories of AI-based network automation are defined in the following in terms of the factors and the features
discussed previously:
• Category 0. Manual O&M: O&M operators manually control the network through traditional interfaces and
check network alarms and logs in order to understand the presence of anomalous behaviours. SNMP events
trigger simple adaptive actions in network devices and alarms provide information about network anomalies
and trigger an action from the operator to solve the problem. The operator is therefore always in the loop.
There is no intelligence in the network though there is some automation needed to activate the single alarms
and the logs triggering the operator's actions.
• Category 1. Assisted O&M: Automated scripts are used in service provisioning, network function
deployment, configuration and maintenance. There is a shallow perception of the network status, limited to a
part of the network or of services, and machines provide limited suggestions for decision making to the
operator. There is local awareness based on alarms, SNMP events and logs, KPI measurement. There is a very
low level of self-adaptation.
ETSI
13 ETSI GR ENI 007 V1.1.1 (2019-11)
• Category 2. Partial automation: Most of the service provisioning is automated, as well as network
deployment and maintenance. The AI system has a comprehensive perception of the network status and local
machine decision making is implemented; the AI system provides multiple options for service provisioning
and can make limited decisions. For example, in cloud computing scenarios with Category 2 network
autonomicity, the data center hosted virtualized part of the network provides the API interface for overlay
network configuration and automatic network configuration operations are performed according to the
scheduling requirements of the cloud platform, such as OpenStack.
• Category 3. Conditional automation: Building on Category 2 capabilities, the AI system can sense real-time
environmental changes, and in certain domains, optimize and adjust the network configuration thanks to the
implementation of closed-loop management. The key characteristic of a Category 3 network is the ability to
perform a dynamic autonomic control of operation and maintain a predefined performance level when
operating in a single domain. In particular, for Category 3, under predefined conditions and in a given domain,
e.g. OSS, BSS, mobile application, transport network, etc., the system can continuously execute control tasks
to assure the target performance. This could be done with the system aggregating alarms and identifying fault
conditions based on AI, triggering the fault location module to quickly locate the fault and automatically
dispatching the trouble ticket. In a Category 3 optical network scenario, for instance, the E2E delay for the
supported services is dynamically measured, traced, optimized and maintained.
• Category 4. High automation: Building on Category 3 capabilities, the AI system enables, in a cross-domain
environment, customer experience-driven predictive or pro-active closed-loop management of networks and
services. This allows operators to resolve network faults prior to receiving customer complaints, reduce service
outages and, ultimately, improve customer satisfaction. The key characteristic of Category 4 automation is the
ability to control and manage the network based on customer experience verification in complex cross-domain
service scenarios. For example, in the home broadband service scenario, in a Category 4 network, there are
autonomic:
- measurement and analysis of customer experience in real time
- continuous identification of dynamic network exceptions with respect to the planned behaviour, and
proactive remediation and improvement to meet or exceed contracted service goals
- rectification of faults and self-recovery
- trouble reporting based on service experience deterioration and network anomalies detection, enabling
predictive operation and maintenance
- resolution of incidents before they escalate (e.g. Event Handling, Self-Healing, Automatic Re-routing,
Resource Reallocation)
- reduction of the service interruption rate and of the related impact on customer experience
• Category 5. Fully autonomic system: This category is the ultimate goal for telecom network evolution. The
system is implemented with full closed-loop automation across multiple services, multiple domains, and the
entire lifecycle, achieving the full autonomicity of the network.
5.4 Tabular representation of Network Autonomicity categories
Table 1 shows how the factors specified in clause 5.2 impact each of the 6 categories of network autonomicity
described in clause 5.3 from a technical point of view. Table 2, instead, analyses the categories in terms of market
perception and impact.
ETSI
14 ETSI GR ENI 007 V1.1.1 (2019-11)
Table 1: Categories of network autonomicity from a technical point of view
Category Name Definition Man- Decision Data Collection Degree of Environment Supported scenarios
Machine Making and Analysis intelligence adaptability
Interface Participation
Category 0 Manual O&M O&M operators manually control How All-manual Single and shallow Lack of AI based Fixed Single scenario
the network and obtain network (command) awareness (SNMP understanding
alarms and logs events and alarms) (manual
management and
control)
Category 1 Assisted O&M Automated scripts are used in How Provide Local awareness Limited analysis Limited adaptability Selected scenarios
service provisioning, network (command) suggestions (SNMP events, capability to changes
deployment, and maintenance. for machines alarms, KPIs, and
Shallow perception of network or humans logs)
status and machine suggestions and help
for decision making decision
making
Category 2 Partial Automation of most service How The machine Comprehensive Deep analysis Limited adaptability Selected scenarios
automation provisioning, network (declarative) provides awareness (basic capability to changes
deployment, and maintenance multiple telemetry data)
Comprehensive perception of opinions, and
network status and local machine the machine
decision making makes limited
decisions
Category 3 Conditional In specific environmental and How Most of the Comprehensive Comprehensive Adaptability to Multiple scenarios
automation network conditions there is (declarative) machines and adaptive analysis and significant changes
automatic network control and make sensing (such as knowledge;
adaptation decisions data compression Short-term forecast
and optimization capability
technologies)
Category 4 Partial Deep awareness of network What Optional Adaptive posture Comprehensive Adaptability to Multiple scenarios
autonomicity status; in most cases the network (intent) decision- awareness analysis and significant changes
performs autonomic making knowledge
decision-making and operation response Long-term forecast
adjustment capability
Category 5 Full autonomicity In all environmental and network What Machine Adaptive Autonomic Adaptability to any Any scenario
conditions, the network can (intent) autonomous optimization as a evolution and change
automatically adapt decision consequence of knowledge
quality of service reasoning
deterioration
ETSI
15 ETSI GR ENI 007 V1.1.1 (2019-11)
Table 2 shows how the factors specified in clause 5.1 impact each of the 6 categories of network autonomicity from a market point of view.
Table 2: Categories of network autonomicity from a market point of view
Category Name Definition Scheduling Perception Analysis and Customer System capability Example of
execution monitoring decision-making experience network generation
Category 0 Manual O&M O&M operators Operator Operator Operator Operator n/a Command line
manually control the
network and obtain
network alarms and logs
Category 1 Assisted O&M Automated scripts are Operator and Operator Operator Operator Selected service scenarios NMS
used in service system
provisioning, network
deployment, and
maintenance. Shallow
perception of network
status and machine
suggestions for decision
making
Category 2 Partial Automation of most Operator and Operator Operator Operator Selected service scenarios NMS + controller
automation service provisioning, System
network deployment,
and maintenance
Comprehensive
perception of network
status and local machine
decision making
Category 3 Conditional In specific environmental Mostly Operator Operator Operator Multiple service scenarios Single-domain:
automation and network conditions System and system Automation + perception analysis +
there is automatic limited context-awareness trigger
network control and conditions drive closed-loop
adaptation management
Category 4 Partial Deep awareness of Mostly Operator Operator Operator and Multiple service scenarios Cross-domain (for some service
autonomicity network status; in most System and System and System System scenarios): Automation +
cases the network perception analysis + experience;
performs autonomic; context-awareness and simple
decision-making and cognitive processing closed-loop
operation adjustment management
Category 5 Full In all environmental and System System System System Any service scenario Cross-domain and any service:
autonomicity network conditions, the Automation + perception analysis +
network can experience; situation awareness
automatically adapt and cognitive processing closed-
loop management
ETSI
16 ETSI GR ENI 007 V1.1.1 (2019-11)
5.5 Scenario Examples of Network Autonomicity categories
5.5.1 Introduction
In order to accelerate adoption of network autonomicity, operators are expected to focus on O&M approaches in the
various scenarios. This means that O&M process, especially in terms of decision making, data collection and analysis,
changes directly relate to a particular goal and result defined by the operator, and with a business value. Progress of
network autonomicity will be accelerated if autonomicity categories of a core set of scenarios is defined, which will be
of value to all operators. Then, development of the related autonomous driving solutions for these scenarios can be
prioritized accordingly. Clause 5.5 illustrates the autonomicity categories of several typical scenarios.
5.5.2 Autonomicity categories of network traffic classification
Network traffic classification works as a fundamental tool in network operation and management, e.g. to ensure QoS,
perform traffic engineering, guarantee network security, etc. With the growth in the diversity of applications, traffic
volume and the proportion of encapsulated traffic, traditional traffic classification methods e.g. the port-based method
and the DPI-based method are inefficient. However applying AI algorithms to the statistic-based method becomes the
main trend.
Table 3 shows the application of autonomicity categories to the network traffic classification scenario.
ETSI
17 ETSI GR ENI 007 V1.1.1 (2019-11)
Table 3: Categories of network autonomicity for network traffic classification
Man-
Decision Making Data Collection and
Category Name Definition Machine Degree of Intelligence Supported Scenarios
Participation Analysis
Interface
Category 0 Manual O&M operators manually How All-manual Manually extracts ports Manually using a port to Non-encrypted traffic
O&M control the network and (command) determine application classification
obtain network alarms and layer protocol
logs
Category 1 Assisted Automated scripts are How Manual analyse the Machines such as DPI Manually discovering Non-encrypted traffic
O&M used in service (command) classification results and devices uses automated which signature strings classification
provisioning, network make decisions such as scripts and inspects match an application
deployment, and Qos guarantee, traffic signatures in the packet
maintenance. Shallow engineering and intrusion payload to identify the traffic
perception of network detection, etc. Then of a non-encrypted
status and machine machines take actions. application
suggestions for decision
making
Category 2 Partial Automation of most HOW Manual analyse the Machines such as DPI Manually discovering Non-encrypted traffic and a
automation service provisioning, (declarative) classification results and devices uses automated which signature strings small amount of encrypted
network deployment, and make decisions such as scripts and inspects match an application traffic classification
maintenance Qos guarantee, traffic signatures in the packet
Comprehensive engineering and intrusion payload to identify the traffic
perception of network detection, etc. Then of a non-encrypted
status and local machine mach
...