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)

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






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.

---------------------- Page: 1 ----------------------
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

---------------------- Page: 2 ----------------------
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

---------------------- Page: 3 ----------------------
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

---------------------- Page: 4 ----------------------
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

---------------------- Page: 5 ----------------------
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

---------------------- Page: 6 ----------------------
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

---------------------- Page: 7 ----------------------
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

---------------------- Page: 8 ----------------------
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

---------------------- Page: 9 ----------------------
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

---------------------- Page: 10 ----------------------
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

---------------------- Page: 11 ----------------------
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 as
...