Professional Documents
Culture Documents
© ETSI 2019
**Experiential Networked Intelligence
Outline
© ETSI 2019 2
ENI Published Deliverables & Workplan
Published ENI deliverables: Ongoing ENI Work Items and Rapporteurs:
• ETSI GR ENI 001 V1.1.1 (2018-04) Published • ENI 001 (WI RGS/ENI-008)
Use Cases – Yue Wang (Samsung) Use Cases (Release 2) – Yue Wang (Samsung)
• ETSI GS ENI 002 V1.1.1 (2018-04) Published • ENI 002 (WI RGS/ENI-007)
Requirements – Haining Wang (China Telecom) Requirements (Release 2) – Haining Wang (China Telecom)
• ETSI GR ENI 003 V1.1.1 (2018-05) Published • ENI 004 (WI RGR/ENI-010)
Context-Aware Policy Management Gap Analysis – John Strassner Terminology (Release 2) – Yu Zeng (China Telecom)
(Huawei)
• ENI 005 (WI DGS/ENI-005)
• ETSI GR ENI 004 V1.1.1 (2018-05) Published System Architecture – John Strassner (Huawei)
Terminology – Yu Zeng (China Telecom)
• ENI 006 (WI RGS/ENI-012)
• ETSI GS ENI 006 V1.1.1 (2018-05) Published PoC Framework (Release 2) – Bill Wright (Redhat/IBM)
Proof of Concept (PoC) Framework - Luca Pesando (TIM)
• ENI 007 (WI RGR/ENI-011)
Definition of Networked Intelligence Categorization – Luca Pesando
(TIM)
© ETSI 2019 3
Business Value
• Network: Traditional SDN & NFV Autonomic Network
SDN & +
NFV
ETSI ISG ENI founded at 17Q1, Release 1 (2017-2019) Role Name (Organization)
• The ISG ENI focuses on improving the operator experience,
adding closed-loop artificial intelligence mechanisms based Chairman Dr. Raymond Forbes (Huawei)
on context-aware, metadata-driven policies to more quickly
recognize and incorporate new and changed knowledge, Vice Chairman Mrs. Haining Wang (China Telecom)
and hence, make actionable decisions. Second Vice Chairman Dr. Luca Pesando (Telecom Italia)
• In particular, ENI will specify a set of use cases, and the
architecture, for a network supervisory assistant system Secretary Dr. Yue Wang (Samsung)
based on the ‘observe-orient-decide-act’ control loop
Technical Officer Mrs. Sylwia Korycinska (ETSI)
model.
• This model can assist decision-making systems, such as Technical Manager Dr. Shucheng Liu “Will” (Huawei)
network control and management systems, to adjust
Raymond Forbes (Huawei)
services and resources offered based on changes in user Sylwia Korycinska (ETSI Technical Officer)
needs, environmental conditions and business goals. Michele Carignani (ETSI CTI)
Extended at 19Q1 into Release 2(2019-2021) ENI ISG PoC Review
Haining Wang (China Telecom)
• New Terms of reference included: implementation, PoC, Team Luca Pesando (Telecom Italia)
plug-tests and open-source relationships Mostafa Essa (Vodafone)
Antonio Gamelas (Portugal Telecom)
© ETSI 2019 Source: https://portal.etsi.org/TBSiteMap/ENI/ListOfENIMembers.aspx
6
ENI Members and Participants
- operators, vendors and research institutes from across Europe, USA and Asia
Officials
PoC Review
Team
Members
Participants
Identify requirements to
improve experience
Standardize how the network
experience is measured
© ETSI 2019 8
Use Cases
Network Operations Network Assurance
Policy-driven IP managed networks Network fault identification and prediction
Radio coverage and capacity optimization Assurance of Service Requirements
Intelligent software rollouts Network Fault Root-cause Analysis and
Policy-based network slicing for IoT security Intelligent Recovery
Intelligent fronthaul management and orchestration
Elastic Resource Management and Orchestration
Application Characteristic based Network Operation Service Orchestration and Management
AI enabled network traffic classification Context aware VoLTE service experience optimization
Intelligent time synchronization of network Intelligent network slicing management
Intelligent carrier-managed SD-WAN
Infrastructure Management Intelligent caching based on prediction of content
Policy-driven IDC traffic steering popularity
Handling of peak planned occurrences
Energy optimization using AI
Source: ETSI RGS/ENI-008 , Experiential Networked Intelligence (ENI); ENI use cases; – Wang, Yue (Samsung)
© ETSI 2019 9
ENI High-Level Functional Architecture
© ETSI 2019 High-Level Functional Architecture Diagram in DGS/ENI-005 (GS ENI 005) v0.0.21 10
Initial Thoughts on Reference Architecture
Functional Architecture with its Input Reference Points Functional Architecture with its Output Reference Points
© ETSI 2019 11
ENI Assisting MANO
Policy
or
• For MANO to take full advantage of ENI, existing interfaces extension or in some cases, new interfaces may be required
© ETSI 2019 • Physical Network interaction, e.g. with SDN Controller explicitly depicted through NFVI interaction (OOB possible too) 12
ENI Assisting MEF LSO RA
Policy
ENI
Autonomous
Adaptive
Automatic
Manual
Control
Use Cases GS ENI 001 Yue Wang Samsung Jul’18 Jul’19 Aug’19 Rel.2 progressing (Rel.1 Published 2018-04-20)
Requirements GS ENI 002 Haining Wang China Telecom Jul’18 Jul’19 Aug’19 Rel.2 progressing (Rel.1*SDNIA:
Published 2018-04-17)
SDN/NFV Industry Alliance,
**SliceNet: H2020 & 5G-PPP Project
Context Aware
GR ENI 003 John Strassner Huawei Sept’17 Dec’17 Mar’18 Rel.1 Published
Policy Modeling
Terminology GR ENI 004 Yu Zeng China Telecom Jul’18 Jul’19 Aug’19 Rel.2 progressing (Rel.1 Published 2018-06-08)
Architecture GS ENI 005 John Strassner Huawei Feb’18 Jul’19 Aug’19 Progressing
PoC Framework GS ENI 006 Bill Wright Redhat/IBM Feb’19 Jun’19 Jul’19 Rel.2 progressing (Rel.1 Published 2018-06-08)
Definition of
Networked Intelligence GR ENI 007 Luca Pesando TIM Oct’18 Jul’19 Aug’19 Progressing
Categorization
© ETSI 2019 15
Categorization of Network Intelligence
Category Name Definition Man- Decision Making Decision Making and Degree of Environment Supported
Machine Participation Analysis Intelligence Adaptability Scenario
Interface
Level 0 Traditional O&M personnel manually control the network How All-manual Single and shallow Lack of Fixed Single
manual and obtain network alarms and logs (comman awareness (SNMP events understanding scenario
network d) and alarms) (manual
Continuous improvement
Autonomous capability
understanding
Level 1 Partially Automated scripts are used in service How Provide suggestions Local awareness (SNMP A small amount Little change Few
automated provisioning, network deployment, and (comman for machines or events, alarms, KPIs, and of analysis scenarios
network maintenance. Shallow perception of network d) humans and help logs)
status and decision making suggestions of decision making
automated machine
diagnostics
Level 2 Automated Automation of most service provisioning, HOW The machine provides Comprehensive awareness Powerful analysis Little change Few
network network deployment, and maintenance (declarativ multiple opinions, and (Telemetry basic data) scenarios
Comprehensive perception of network status e) the machine makes a
and local machine decision making small decision
Level 3 Self- Deep awareness of network status and HOW Most of the machines Comprehensive and Comprehensive Changeable Multiple
optimization automatic network control, meeting users' (declarativ make decisions adaptive sensing (such as knowledge scenarios and
network network intentions e) data compression and combinations
optimization technologies) Forecast
Level 4 Partial In a limited environment, people do not need to WHAT Optional decision- Adaptive posture awareness Comprehensive Changeable Multiple
autonomous participate in decision-making and adjust (intent) making response (edge collection + knowledge scenarios and
network themselves (decision comments of judgment) combinations
the challenger) Forward forecast
Level 5 Autonomou In different network environments and network WHAT Machine self-decision Adaptive deterioration Self-evolution Any change Any scenario
s network conditions, the network can automatically adapt (intent) optimization (edge closed- and knowledge &
to and adjust to meet people's intentions loop, including collection, reasoning combination
© ETSI 2019 judgment, and optimization) 16
Example of categorization report
Work Plan
Release 2: Part1
Tasks:
Release 1: Part2 • Network Intelligence Categorization
Release 1: Part1 Tasks: • System Architecture
Tasks: • Reference architecture • Data model
• Use Cases • PoCs demonstrating different scenarios • Interface/ external Reference points
• Requirements • Updated Terminology, Use Cases, and • Security and Validation specification
• Gap analysis Requirements • Implementation:
• Terminology • PoCs demonstrations
Output: • Plug-tests
Output: • Group Reports and/or • Open-source
• Group Reports showing cross- SDO Normative Group Specifications based on phase • Categorization, Terminology, Use Cases, and
•Operator reqs functional architecture, interfaces/APIs, and 1 study Requirements
•ENI concepts models or protocols, addressing stated Output:
requirements • Normative Group Specifications
• Group Reports as required
Dec’16 Feb’17 Apr’17 May’17 Sep&Oct’17 Dec’17 Mar’18 Jun’18 Sep’18 Dec’18 Apr’19 July ’19 Sept ’19 Dec’ 19 … 2021
Brainstorm ETSI ISG Kickoff ENI#03 ENI#04, ENI#05 ENI#06 ENI#07 ENI#08 ENI#09 ENI#10 ENI#11 ENI#12
ENI ENI#02
event meeting China Telecom Samsung ETSI TIM China Telecom 5Tonic/Telefonica Samsung Portugal Telecom China Telecom ETSI(?)
created ETSI
ETSI Beijing, London, Turin Beijing Madrid Warsaw Aveiro Beijing
SNDIA joint SliceNet joint
workshop, workshop
Whitepaper
published
Release 1 focus on use case & requirements, started designing architecture and PoCs.
The ETSI Director-General advised by the ETSI Board approved in Feb. 2019 that ISG ENI is mandated to work on ENI release 2 for 2019-2021.
Four F2F meetings per year: 19Q2 Hosted by Portugal Telecom in Aveiro, 19Q3 Hosted by China Telecom in Beijing.
© ETSI 2019 17
Online meeting every week https://portal.etsi.org/tb.aspx?tbid=857&SubTB=857#5069-meetings
PoC Team and ENI Work-Flow proposal
using the process under definition in ETSI
Procedures:
ISG ENI draft and approve a PoC framework PoC team PoC review ISG ENI
team
Form a PoC review group to receive and review PoC proposals with 2 1
PoC topics PoC topic list New PoC New PoC topic
formal delegation from ISG list maintenance topic identification
PoC teams (the proposers – which may include non-members) shall Comments No
Accepted?
present an initial proposal and a final report, according to the
templates given by ISG for review 5
Yes
PoC
start
PoC Team(s) are independent of the ISG – use the process and 6
7
template of the ISG – Choose a POC Team Leader and draft the PoC project
PoC project contributions
PoC
Feedback contributions
proposal lifetime
handling
PoC
Report
8
e.g.
ENI PoC review team: PoC report
review contributions
to the system
Architecture
Comments No Accepted
?
Ye
9 s
PoC end
PoC #5: AI Enabled Network China Mobile Weiyuan Li ENI(19)009_ Proposed Proposed
Traffic Classification Huawei, Intel, Tsinghua University 004r5 in Apr
2019
TNSM:
• Provides underlay network
control to satisfy the network
slice requests.
• FlexE and a FlexE-based
optimization algorithm are used
for underlay network slice
creation and modification.
CNSM:
• Provides core network control
to satisfy the network slice
requests
✓ PoC Project Goal #1: Demonstrate the use of AI to predict the change of traffic pattern and adjust
the configuration of network slice in advance.
✓PoC Project Goal #2: Demonstrate the use of intent based interface to translate tenant requirements
https://eniwiki.etsi.org/index.php?title=PoC#PoC.231
to network slice configuration and intelligent network slice lifecycle management on demand.
© ETSI 2019 :_Intelligent_Network_Slice_Lifecycle_Management 20
ENI PoC project #2: Elastic Network Slice Management
Main Features One innovative service
through 2 network slices
Network Slice Blueprinting
& Onboarding Virtual Reality application
© ETSI 2019 Hardware and Software Setup Design, test and validate new interfaces with MANO 21
https://eniwiki.etsi.org/index.php?title=PoC#PoC.232:_Elastic_Network_Slice_Management
ENI PoC project #3: Securing against Intruders and other threats
through a NFV-enabled Environment (SHIELD)
Status: PoC public demo 23-25th Feb. 2019, and finished.
Host/Team Leader:
Team members::
https://www.shield-h2020.eu/
Trust Monitor
© ETSI 2019 22
https://eniwiki.etsi.org/index.php?title=PoC#PoC.233:_Securing_against_Intruders_and_other_threats_through_a_NFV-enabled_Environment
ENI PoC project #4: Predictive Fault management of E2E
Multi-domain Network Slices
Energy Distributor https://slicenet.eu/
(Vertical)
Network Service Provider (NSP) A Network Service Provider (NSP) B PoC Project Goal #1: Network Slice Fault Prediction.
Demonstrate the use of AI on performance data to
RAN EPC
Core
RAN be able to accurately predict failure situations on
IP/MPLS
Network Slices and estimate their impact on an E2E
multi-domain slice performance.
https://eniwiki.etsi.org/index.php?title=PoC#PoC.234:_Predictive_Fault_management_of_E2E_Multi-domain_Network_Slices
© ETSI 2019 23
Ecosystem
Open Source Standard & Industry Research
IRTF NMRG
Linux Foundation
ONAP, Acumos, PNDA ETSI ENI
(Data collection and decision, AI ) ML5G
EU H2020 Projects
ITU-T, MEF, TMF CCSA TC 610 - AIAN
(Policy, Models, Architecture) (AI applied to Network)
• Cooperate with mainstream operators, vendors and research institutes in Europe, USA and Asia
• Collaborate with other SDOs and industry ad-hocs
• Liaisons exchanged with IETF, BBF, MEF, ITU-T
• Liaisons with other ETSI groups: NFV, NGP, MEC, NTECH, OSM, ZSM
• Position ETSI ENI as the home of network intelligence standards
• Guide the industry with consensus on evolution of network intelligence
© ETSI 2019 • Boarder between different categories are becoming vague. 24
Network Intelligence Activities in 2016 - 2019
Forum on Network Intelligence, Dec’16, Shenzhen, China
• Orange, NTT, CableLabs, China Telecom, China Unicom, China Mobile, NEC, HPE, HKUST etc, 40+ participants joined
• Study of the scope of network intelligence and related technologies, collected ideas of operators and academics
• Participants discussed and get rough consensus on the ENI ISG Proposal
ENI & SDNIA Joint Forum on Network Intelligence, Sep’17, Beijing, China
• China Telecom, China Unicom, China Mobile, Huawei, Nokia, Intel, Samsung, ZTE, H3C, Lenovo, HPE, etc , 140+ participants joined Forum on Network Intelligence, Dec’16
• ENI main players presented the progress of ENI
• Operators shared use cases and practices
• Manufacturers shared solution ideas
• Demo prototype that embodies the ENI concept
• Deep cooperation between SDNIA and ENI was discussed, after that SDNIA created new industry group AIAN(AI Applied to
Network)
ENI & 5GPPP MoNArch Workshop, June’18, Turin Italy ENI & SDNIA Joint Forum on Network
Intelligence, Sep’17
• Collaboration on PoC and Common partners identified between ENI and 5G=MoNArch
• Need to Formally submit the PoC Proposal
ENI & CCSA TC 610 ANIA Joint Forum on Network Intelligence, Sep’18, Beijing, China
• ENI main players presented the progress of ENI
• Operators shared PoC experience and demos with support of Manufacturers
• Demo prototype that embodies the ENI concept
ENI & SliceNet workshop, Dec’17
© ETSI 2019 25
Other related SDOs, Industry Consortia and EU Projects
ORGANIZATION ACTIVITY
ITU-T SG13 FG ML5G Focus Group on Machine Learning for Future Networks including 5G
IETF/IRTF ANIMAWG , IETF Working Group on Autonomic Networking Integrated Model and Approach
NMRG IRTF Research Group on Network Management
CCSA TC610 (was SDNIA) AIAN (Artificial Intelligence Applied in Network) industry group
© ETSI 2019 26
Other ETSI internal Technical Bodies (TCs/ISGs)
ETSI TECHNICAL BODY ACTIVITY
ETSI ISG ZSM Industry Standardization Group on Zero touch network and Service Management
ETSI TC INT - AFI Technical Committee Core Network and Interoperability Testing - Evolution of Management
towards Autonomic Future Internet
Contact Details:
Chair: Dr. Raymond Forbes
rayforbes@mail01.huawei.com
Thank you!
© ETSI 2019 28
ENI
Use Cases
© ETSI 2019
Overview of the Use Case Work Item
Identify and describe appropriate use cases
© ETSI 2019 30
Use Cases
Network Operations Network Assurance
Policy-driven IP managed networks Network fault identification and prediction
Radio coverage and capacity optimization Assurance of Service Requirements
Intelligent software rollouts Network Fault Root-cause Analysis and
Policy-based network slicing for IoT security Intelligent Recovery
Intelligent fronthaul management and orchestration
Elastic Resource Management and Orchestration
Application Characteristic based Network Operation Service Orchestration and Management
AI enabled network traffic classification Context aware VoLTE service experience optimization
Intelligent network slicing management
Infrastructure Management Intelligent carrier-managed SD-WAN
Policy-driven IDC traffic steering Intelligent caching based on prediction of content
Handling of peak planned occurrences popularity
Energy optimization using AI
Source: ETSI RGS/ENI-008 , Experiential Networked Intelligence (ENI); ENI use cases; – Wang, Yue (Samsung) = examples shown in the next slides
© ETSI 2019 31
ENI Use Case:
Policy-driven IDC traffic steering
IDC 1 IDC 2
Link 2: light loaded
The challenges:
➢ Multiple links between IDCs, allocated by network administrator, e.g., shortest path strategy
➢ Link load not sufficiently considered when calculating the traffic path
➢ Bandwidth allocated to a tenant is not always fully used all time
The challenges:
➢ The servers in a DC take 70% of the total power consumption
➢ Servers are deployed and running to meet the requirement of peak hour service –
100% powered-up full time.
The challenges:
➢ Massive deployment of devices
➢ Slices may be dynamically expansible and adaptable in a changing context
➢ Methods and technologies of attacks are widely changing
The role of AI and ENI:
➢ Signal specific traffic patterns indicating Distributed Denial of Service (DDOS) attack or other
type of attacks
➢ Automatically isolate the attacked devices
© ETSI 2019 34
ENI Use Case:
Intelligent Fronthaul Management and Orchestration
PDCP
RLC
RAU
High MAC
Fronthaul Orchestration
Cluster
and Management
Low MAC
High PHY
Low PHY
RF
Fronthaul RCC
The challenges:
➢ Network traffic classification can support numerous network closed-loop control in terms of network security, traffic
engineering, Quality of Service (QoS)
➢ By providing traffic change or distribution information at network or service level, classification results support policy-making
processes and sever as the guideline for Operations support system (OSS) and traffic forecast.
The role of AI and ENI:
➢ Current classification methods cannot sufficiently (or fail to) support the growth in the diversity of applications, traffic volume
and the proportion of encapsulated traffic
➢ AI enabled Network Traffic classifier transforms bit streams or extracted features into images and models the network traffic
classification as the ‘traffic image’ classification.
➢ By collecting data at the transport level (instead of application level) and applying machine learning algorithms, the ENI system
© ETSIis2019
able to achieve accurate classification result with a relatively low overhead. 36
ENI Use Case:
Intelligent Network Slicing Management
SMF UPF …
Common NFs
The challenges:
➢ Dynamic traffic
➢ Allocation of network resources intra/inter slices (VNF scale in/out, up/down)
The role of AI and ENI:
➢ Analyze collected data associated to e.g., traffic load, service characteristic, VNF type and
constraints, infrastructure capability and resource usage, etc.
➢ Produce a proper context aware policy to indicate to the network slice management entity
when, where and how to place or adjust the network slice instance (e.g., reconfiguration, scale-
in, scale-out, change the template of the network slice instance)
© ETSI 2019 ➢ Dynamically change a given slice resource reservation 37
ENI Use Case:
Assurance of Service Requirements
The challenges:
➢ Dedicated slice (e.g., banking, energy provider company)
➢ Vital applications of the network to operate core business
➢ Very strict requirements (e.g., a personalized SLA)
Thank you!
© ETSI 2019 40
© ETSI 2019. All rights reserved
ENI
Requirements
and
Terminology
Haining Wang (China Telecom), Rapporteur, Requirements
© ETSI 2019
Overview of ENI requirements Work Item
General information:
Creation Date: 2018-04-20 Type: Group Specification
Work Item RGS/ENI-007 Latest version: 2.0.4
Reference:
Rapporteur: Haining Wang Technical Officer: Sylwia Korycinska
Scope:
This document specifies the requirements related to ENI. This includes the
common requirements for both fixed and mobile, fixed specific requirements
and mobile specific requirements. The ENI requirements are based on the ENI
use case document and identified requirements from other SDOs. These
requirements will form the base for the architecture design work. This revision
will update the existing requirements where needed and add new functional
and non-functional, security and architectural requirements.
© ETSI 2019 42
Categorization of the Requirements
Level 1 Level 2
Functional Data Collection and Analysis
requirements Policy Management
Level 1 Level 2 Data Learning
Service and network General requirements Interworking with Other Systems
requirements Service orchestration and management Mode of Operations
Network planning and deployment
Level 1 Level 2
Network optimization
Non-functional Performance requirements
Resilience and reliability
requirements Operational requirements
Security and privacy
Regulatory requirements
Non-functional policy
requirements
© ETSI 2019 43
Overview of ENI Work Item - Terminology
General information:
Creation Date: 2018-06-29 Type: Group Report
Work Item RGR/ENI-010 Latest version: 2.0.1
Reference:
Rapporteur: Yu Zeng Technical Officer: Sylwia Korycinska
Scope:
The WI will provide terms and definitions used within the scope of the ISG ENI, in
order to achieve a "common language" across all the ISG ENI documentation. This
work item will be updated with the general terminology required as the ENI
specifications develop.
© ETSI 2019 44
Work Plan and Next Steps
Thank you!
© ETSI 2019 45
ENI
Intelligent
Policy
Management
Dr. John Strassner, (Huawei), Rapporteur, Context-Aware Policy Management
© ETSI 2019
Overview of ENI Work Item – Context Aware Gap-Analysis
General information:
Creation Date: 2017-04-10 Type: Group Report
Work Item DGR/ENI-003 Latest version: 1.1.1
Reference:
Rapporteur: John Strassner Technical Officer: Sylwia Korycinska
Scope:
A critical foundation of experiential networked intelligence is context- and
situation-awareness. This WI will analyze work done in various SDOs on
network policy management in general, and context-aware network policy
management specifically, to determine what can be reused and what needs to
be developed. The requirements documented in this report will be considered
during the architecture design work.
© ETSI 2019 47
Main Contents of the Context-Aware Policy Management WI
Content
Introduction and Approach
• Including Introduction to Policy Management, The Policy
Continuum, Types of Policy Paradigms, etc.
The number of continua in the Policy Continuum should be determined by the applications using it.
There is no fixed number of continua.
© ETSI 2019 49
Policy Examples: Intent
Intent policies describe the set of computations that need to be
done without describing how to execute them
Policies are written in a restricted natural language
The flow of control is not specified
The order in which operations occur are irrelevant
Each statement in an Intent Policy may require the translation
of one or more of its terms to a form that another managed
functional entity can understand.
Express What should be done,
not How to do it.
© ETSI 2019 Intent policy is an ongoing work in MEF, ENI may reuse some of the output 50
Work Plan and Next Steps
Thank you!
© ETSI 2019 51
Experiential
Networked
Intelligence
System
Architecture
Dr. John Strassner (Huawei, Rapporteur)
© ETSI 2019
Overview of ENI Work Item – System Architecture
General information:
Creation Date: 2017-12-12 Type: Group Specification
Work Item DGS/ENI-005 Latest version: 0.0.21
Reference:
Rapporteur: John Strassner Technical Officer: Sylwia Korycinska
Scope:
The purpose of this work item is to define the software functional architecture of ENI. This includes:
• defining the functions and interactions that satisfy the ENI Requirements
• defining an architecture, in terms of functional blocks, that can meet the needs specified by the ENI Use Cases
• defining Reference Points that the above functional blocks use for all communication with systems and entities
© ETSI 2019 53
High-Level Functional System Architecture
© ETSI 2019 54
Closed Loop Control in FOCALE
1. Outer Closed Control Loop for a Given
Context and Long-Term Optimization
2. Inner Closed Control Loop Triggered by
Context Change
Monitor
Design Orchestrate
AI Guidance,
Policy Enforcement
using Model-Driven
AI Guidance, Engineering
Policy Enforcement
using Model-Driven
Engineering
Real-time
Optimize Offline Analytics Analytics
FOCALE Model-Driven Development (1)
FOCALE was designed for
Autonomic Policy Management
FOCALE abstracts the functionality, structure, and behavior of the
system being managed using a unique combination of models,
ontologies, and logic
Context selects policies, which define behavior; as context
changes, policies change
Input state is extracted/inferred from OAMP data and context and
compared to desired state
If they are equal, continue monitoring
© ETSI 2019 58
Context-Aware Management
Functional Block
The purpose of the Context-Aware Management Functional Block is to describe the state
and environment in which an entity exists or has existed. Context consists of measured
and inferred knowledge, and may change over time. For example, a company may have
a business rule that prevents any user from accessing the code server unless that user
is connected using the company intranet. This business rule is context-dependent, and
the system is required to detect the type of connection of a user, and adjust access
privileges of that user dynamically.
© ETSI 2019 59
Situation-Awareness Management
Functional Block
The purpose of the Situation Awareness Functional Block is for the ENI system to be
aware of events and behaviour that are relevant to the environment of the system that it
is managing or assisting. This includes the ability to understand how information, events,
and recommended commands given by the ENI system will impact the management and
operational goals and behavior, both immediately and in the near future. Situation
awareness is especially important in environments where the information flow is high,
and poor decisions may lead to serious consequences (e.g., violation of SLAs).
The working definition of situation awareness for ENI is:
The perception of data and behaviour that pertain to the relevant circumstances and/or
conditions of a system or process (“the situation”), the comprehension of the meaning
and significance of these data and behaviours, and how processes, actions, and new
situations inferred from these data and processes are likely to evolve in the near future
to enable more accurate and fruitful decision-making.
© ETSI 2019 60
Policy-based Management
Functional Block
The purpose of the Policy Management Functional Block is to provide decisions to ensure that the system goals
and objectives are met. Formally, the definition of policy is:
Policy is a set of rules that is used to manage and control the changing and/or maintaining of the state of one or
more managed objects.
There are three different types of policies that are defined for an ENI system:
• Imperative policy: a type of policy that uses statements to explicitly change the state of a set of targeted objects. Hence, the
order of statements that make up the policy is explicitly defined.
In this document, Imperative Policy will refer to policies that are made up of Events, Conditions, and Actions.
• Declarative policy: a type of policy that uses statements to express the goals of the policy, but not how to accomplish those
goals. Hence, state is not explicitly manipulated, and the order of statements that make up the policy is irrelevant.
In this document, Declarative Policy will refer to policies that execute as theories of a formal logic.
• Intent policy: a type of policy that uses statements to express the goals of the policy, but not how to accomplish those goals.
Each statement in an Intent Policy may require the translation of one or more of its terms to a form that another managed
functional entity can understand.
In this document, Intent Policy will refer to policies that do not execute as theories of a formal logic. They typically are
expressed in a restricted natural language, and require a mapping to a form understandable by other managed functional
entities.
An ENI system MAY use any combination of imperative, declarative, and intent policy to express recommendations
and commands to be issued to the system that it is assisting.
© ETSI 2019 61
Cognition Management
Functional Block
Cognition, as defined in the Oxford English Dictionary, is “the mental action or process of acquiring knowledge and
understanding through thought, experience, and the senses”.
The purpose of the Cognition Framework Functional Block is to enable the ENI system to understand ingested data
and information, as well as the context that defines how those data were produced; once that understanding is
achieved, the cognition framework functional block then provides the following functions:
• change existing knowledge and/or add new knowledge corresponding to those data and information
• perform inferences about the ingested information and data to generate new knowledge
• use raw data, inferences, and/or historical data to understand what is happening in a particular context, why
the data were generated, and which entities could be affected
• determine if any new actions should be taken to ensure that the goals and objectives of the system will be met.
A cognition framework uses multiple diverse processes and technologies, including linguistics, computer science,
AI, formal logic, neuroscience, psychology, and philosophy, along with others, to analyse existing knowledge and
synthesise new knowledge
© ETSI 2019 63
© ETSI 2019. All rights reserved
ENI
PoC
Framework
© ETSI 2019
PoC Team and ENI Work-Flow proposal
using the process under definition in ETSI
Procedures:
ISG ENI draft and approve a PoC framework PoC team PoC review ISG ENI
team
Form a PoC review group to receive and review PoC proposals with 2 1
PoC topics PoC topic list New PoC New PoC topic
formal delegation from ISG list maintenance topic identification
PoC teams (the proposers – which may include non-members) shall Comments No
Accepted?
present an initial proposal and a final report, according to the
templates given by ISG for review 5
Yes
PoC
start
PoC Team(s) are independent of the ISG – use the process and 6
7
template of the ISG – Choose a POC Team Leader and draft the PoC project
PoC project contributions
PoC
Feedback contributions
proposal lifetime
handling
PoC
Report
8
e.g.
ENI PoC review team: PoC report
review contributions
to the system
Architecture
Comments No Accepted
?
Ye
9 s
PoC end
PoC #5: AI Enabled Network China Mobile Weiyuan Li ENI(19)009_ Proposed Proposed
Traffic Classification Huawei, Intel, Tsinghua University 004r5 in Apr
2019
TNSM:
• Provides underlay network
control to satisfy the network
slice requests.
• FlexE and a FlexE-based
optimization algorithm are used
for underlay network slice
creation and modification.
CNSM:
• Provides core network control
to satisfy the network slice
requests
✓ PoC Project Goal #1: Demonstrate the use of AI to predict the change of traffic pattern and adjust
the configuration of network slice in advance.
✓PoC Project Goal #2: Demonstrate the use of intent based interface to translate tenant requirements
https://eniwiki.etsi.org/index.php?title=PoC#PoC.231
to network slice configuration and intelligent network slice lifecycle management on demand.
© ETSI 2019 :_Intelligent_Network_Slice_Lifecycle_Management 67
ENI PoC project #2: Elastic Network Slice Management
Main Features One innovative service
through 2 network slices
Network Slice Blueprinting
& Onboarding Virtual Reality application
© ETSI 2019 Hardware and Software Setup Design, test and validate new interfaces with MANO 68
https://eniwiki.etsi.org/index.php?title=PoC#PoC.232:_Elastic_Network_Slice_Management
ENI PoC project #3: Securing against Intruders and other threats
through a NFV-enabled Environment (SHIELD)
Status: PoC public demo 23-25th Feb. 2019, and finished.
Host/Team Leader:
Team members::
https://www.shield-h2020.eu/
Trust Monitor
© ETSI 2019 69
https://eniwiki.etsi.org/index.php?title=PoC#PoC.233:_Securing_against_Intruders_and_other_threats_through_a_NFV-enabled_Environment
ENI PoC project #4: Predictive Fault management of E2E
Multi-domain Network Slices
Energy Distributor https://slicenet.eu/
(Vertical)
Network Service Provider (NSP) A Network Service Provider (NSP) B PoC Project Goal #1: Network Slice Fault Prediction.
Demonstrate the use of AI on performance data to
RAN EPC
Core
RAN be able to accurately predict failure situations on
IP/MPLS
Network Slices and estimate their impact on an E2E
multi-domain slice performance.
https://eniwiki.etsi.org/index.php?title=PoC#PoC.234:_Predictive_Fault_management_of_E2E_Multi-domain_Network_Slices
© ETSI 2019 70
Definition of
Networked
Intelligence
Categorization
© ETSI 2019
ENI 007 Overview of ENI Work Item – Definition of Network
Networked Intelligence Categorization
General information:
Creation Date: 2018-9-19 Type: Group Report (GR)
Work Item DGR/ENI-0011 Latest version: V0.0.8
Reference:
Rapporteur: Luca Pesando Technical Officer: Sylwia Korycinska
Scope:
This Work Item will address the aspects of gradual implementation of networked intelligence and specify a
categorization framework for systems based on several different levels (e.g. 5, ranging from fully manual to fully
autonomic systems as written in ENI whitepaper). This categorization framework will be based on the amount of
network engineer intervention and attentiveness required and the expected interfaces with the assisted system.
The details of each level will be defined according to scenarios supported and techniques needed.
© ETSI 2019 72
Categorization of Network Intelligence
Category Name Definition Man- Decision Making Decision Making and Degree of Environment Supported
Machine Participation Analysis Intelligence Adaptability Scenario
Interface
Level 0 Traditional O&M personnel manually control the network How All-manual Single and shallow Lack of Fixed Single
manual and obtain network alarms and logs (comman awareness (SNMP events understanding scenario
network d) and alarms) (manual
Continuous improvement
Autonomous capability
understanding
Level 1 Partially Automated scripts are used in service How Provide suggestions Local awareness (SNMP A small amount Little change Few
automated provisioning, network deployment, and (comman for machines or events, alarms, KPIs, and of analysis scenarios
network maintenance. Shallow perception of network d) humans and help logs)
status and decision making suggestions of decision making
automated machine
diagnostics
Level 2 Automated Automation of most service provisioning, HOW The machine provides Comprehensive awareness Powerful analysis Little change Few
network network deployment, and maintenance (declarativ multiple opinions, and (Telemetry basic data) scenarios
Comprehensive perception of network status e) the machine makes a
and local machine decision making small decision
Level 3 Self- Deep awareness of network status and HOW Most of the machines Comprehensive and Comprehensive Changeable Multiple
optimization automatic network control, meeting users' (declarativ make decisions adaptive sensing (such as knowledge scenarios and
network network intentions e) data compression and combinations
optimization technologies) Forecast
Level 4 Partial In a limited environment, people do not need to WHAT Optional decision- Adaptive posture awareness Comprehensive Changeable Multiple
autonomous participate in decision-making and adjust (intent) making response (edge collection + knowledge scenarios and
network themselves (decision comments of judgment) combinations
the challenger) Forward forecast
Level 5 Autonomou In different network environments and network WHAT Machine self-decision Adaptive deterioration Self-evolution Any change Any scenario
s network conditions, the network can automatically adapt (intent) optimization (edge closed- and knowledge &
to and adjust to meet people's intentions loop, including collection, reasoning combination
© ETSI 2019 judgment, and optimization) 73
Example of categorization report
Please Contribute
ETSI ENI#10 meeting will be hosted by Portugal Telecom in Aveiro,
Portugal, on Jul 9-11, 2019.
You are welcome to join us!
Contact Details:
Chair: Dr. Raymond Forbes raymond.forbes@huawei.com
+44 771 851 1361
Future expected issues:
Release 2 on the Categorization, Use cases & requirements
– New Use Cases
– Priority Use Cases demonstrated in PoCs
Development of the system architecture
– data models and APIs
– Interface definitions and External reference points
Establishment of Implementations
– ENI PoC Framework, Contribution to PoC,
– assurance of PoC, Plug-tests, Open Source &
– Validation of the ENI System Architecture
Standardize how the network experience is measured
Welcome new members: especially to be active in discussions
© ETSI 2019
Thank you! 74