ICOIN 2024 Tutorial 3

End-to-End Network Slice

Orchestration and Automation

Nakjung Choi
Head of Mobile Network Systems Department
Network Systems and Security Research
Bell Labs Core Research, Nokia
January 19th, 2024
1 © 2024 Nokia
Agenda
Part I. End-to-End Network Slicing

Part II. Network Slicing Orchestration

Concluding Remarks

2 © 2024 Nokia
 Part I. End-to-End Network Slicing

3 © 2024 Nokia
Diversity of Emerging Services
A wide variety of new use cases and business models

Emerging Use Case

Use Case require diverse network characteristics

Networking Requirements
Data rate Data rate Device
Latency Reliability Mobility Coverage Security
DL UL Density
Optimized networks
Mobile Video Streaming (4K) 15Mbps 500kbps <10k/km² <1s Low Medium Country Low are needed for
Video Surveillance (static, 4K) 300kbps 10Mbps 500/km² <1s Low None City Medium specific use case
characteristics!
Mobile VR/AR Cloud Gaming (4K) 15Mbps 1Mbps <10k/km² <10ms Medium Medium Country Low

V2X Platooning 50Mbps 10Mbps 250/km² <20ms High High Road/Global Critical

IIoT Motion Control 1-10Mbps 1-10Mbps 100k/km² 1ms sync. Very High Low Campus Critical

Tracking and Tracing 100kbps 250kbps 1k/km² >1s Low Medium Global Medium

Optimization focus for use case

4 © 2024 Nokia
Concept of Network Slice
Running Multiple Logical Networks on a Common Physical Infrastructure
Network as a Service: as-a-Service (business) model to
provide flexible consumption of network infrastructure,
including network hardware, software, management
tools, licenses, and life-cycle services
Cloud Scalability and Efficiency
Autonomous
driving
Flexibility to meet Self service
diverse requirements Utility

Automotive

Health
Health
Smart Meter
Software Defined Full Automation and Self-optimization
Networking Distributed Clouds
Network Function
Programmable Network Fabric Virtualization

▪ “Network Slice” is a logical network that provides specific network capabilities and network
characteristics, supporting various service properties for network slice customers
▪ Slicing across access, transport, core, edge and central clouds
5 © 2024 Nokia
5G Network Slice
Slice Type and Slice Identification
NSSAI (Network Slice Selection Assistance Information): A collection of S-NSSAIs, ▪ The SST field may have standardized and
One UE can signal max 8 S-NSSAI in NSSAI non-standardized values
▪ SD makes slice as operator specific
S-NSSAI (Single NSSAI): Identifies single Network Slice, consisting of:
▪ SD is used to allow operator specific handling
▪ A Slice/Service Type (SST: 8 bits): defines slice characteristics, mandatory
of slicing (e.g., different allowed areas, mobile
▪ A Slice Differentiator (SD: 24 bits): identifies slices for more than one slice with
vs FWA service, different changing in 5GC
the same SST, optional

TS 23.501 Table 5.15.2.2-1 - Standardised SST values

Slice/Service
SST value Characteristics. NSSAI
type MBB Slice MNO
eMBB 1 Slice suitable for the handling of 5G eMBB S-NSSAI#1 SST#1 SD#A
URLLC 2 Slice suitable for the handling of URLLC
S-NSSAI#2 SST#1 SD#B
MIoT 3 Slice suitable for the handling of massive IoT
S-NSSAI#3 SST#2 SD#C MBB Slice ENTERPRISE B
V2X 4 Slice suitable for the handling of V2X services
Slice suitable for the handling of High-Performance
HMTC 5
Machine-Type Communications S-NSSAI#8
Slice suitable for the handling of High Data rate URLLC Slice ENTERPRISE C
HDLLC 6
and Low Latency Communications
Standardized SST range (next standardized slices
Standardized 7…127
under definition in 3GPP)
Operator 128..255 Operator specific range

6 © 2024 Nokia
 RRH: Remote Radio Head
RU: Radio Unit

End-to-End Network Slice DU: Distributed Unit

CU: Central Unit
UP: User Plane

Reference Architecture CP: Control Plane

End-to-End Network Slices

E2E NS :=
∑ Domain NSn RAN Transport Core
Slices Slices Slices

Actual domain slice realization is depending

Enterprise-1
on various deployment options
Infotainment
HD maps
Enterprise-1 Enterprise-1
Automated driving infotainment HD maps

Enterprise-2
Automated driving
Transport network

Public and private clouds

Public Safety 5G RAN 5G core

Operator-x
Video surveillance
Multi-source PS video
E2E network slices = NS1 NS2 NS3 NS4 NS5 artificial surveillance
Tenants intelligence

Distributed Clouds

7 © 2024 Nokia
Domain Network Slice
RAN Slice (1/2)
How can RAN Slice be realized?

Geographical Scheduling Local Processing & Routing Cloud Resources

Network Slice ID can be
Network Slice is configured per Network Slice ID can be used for Network Slice ID can be used for
used for allocating
TA (Tracking Area) radio resource scheduling local processing / routing decision
dedicated UP resources
UPF+local application
TA 2-Cell 1 User plane resources
Slice #3 Common Slice #2 Slice #3
TA 1-Cell 1
Slice Slice #1
Slice Resource BBU
#1,2,3
TA 1-Cell 2 Scheduling
TA 2-Cell 2
Slice
Slice #3
#1,2,3
TA 1-Cell 3 Slice to transport mapping
Slice
#1,2,3
DU
CU
3GPP TS 28.541 The 3GPP Network Resource Model (NRM)
• 3 types of resources: Dedicated Resource (DR) ,
Prioritized Resource (PR) , Shared Resource (SR)
• DR exclusively for assigned slice
• Unused PR can be shared across slices
• Sum of three types can be greater than total resources per
gNB
8 © 2024 Nokia
Domain Network Slice
RAN Slice (2/2)

Example: eMBB Slice Example: eMTC Slice Example: URLLC Slice

Layer
(Video streaming) (Temperature Sensor) (Smart Grid)

State handling optimized for Handover measurement Optimized for latency

RRC
RAN/CN signaling omitted reduction

Ciphering omission, header ciphering omission, header

PDCP Default
compression compression

RLC Default Unacknowledged Mode Acknowledged Mode

HARQ optimized for HARQ optimized for high

MAC Default
coverage reliability, RACH priority

Coding optimized for larger Coding optimized for Coding optimized for low
PHY
payload coverage, energy efficiency latency, short payload

Example RAN configuration for different slices

9 © 2024 Nokia
 AMF: Access & Mobility Management Function UDSF: Unstructured Data Storage Function
SMF: Session Management Function NRF: Network Repository Function
UPF: User Plane Function NSSF: Network Slice Selection Function

Domain Network Slice N3IWF: Non-3GPP Interworking Function

PCF: Policy Control Function
AUSF: Authentication Server Function
AF: Application Function
NEF: Network Exposure Function
LMF: Location Management Function

Core Slice
UDM: Unified Data Management SEPP: Security Edge Protection Proxy
UDR: Unified Data Repository NWDAF: Network Data Analytics Function

How can Core Slice be realized?

Core Slice s based on cloud installation

Shared/Dedicated Core network is created in cloud for a given slice
For example, 5G Core selection is based on PLMN and S-NSSAI

Diverse Core NFs:

AMF, SMF, UPF, N3IWF, PCF, AUSF,

UDM, UDR, UDSF, NRF, NSSF, AF,
NEF, LMF, SEPP, NWDAF, …

Private/Public Cloud:

On-premise, far-edge, edge,

metro, central, …

10 © 2024 Nokia
Domain Network Slice
Transport Slice
How can Transport Slice be realized?

A set of Transport Slices Slice Isolation (soft vs. hard) SLA, QoS Standardization

Set of distinct connections VPN Tunnel Creation for dedicated Deterministic SLA Not handled by 3GPP
Realized by any technology (IP, transport resource provisioning - Any QoS Level Marking (e.g. DSCP Multiple standardization organization
Optics, Microwave and PON) and tunnel types (IP, MPLS, SR,…) marking, Queue Size, PIR/CIR etc) working on it: IETF, BBF, MEF
at any layer L0/L1/L2/L3 SDN based solution

From, 3GPP Slice NRM, EP_Transport

instance includes additional attributes
as follow:

▪ ipAddress
▪ localLogicalInterfaceInfo
▪ qosProfile
▪ epApplicationRef
▪ externalEndPointRefList

Multiple transport slices per E2E slice, each having multiple connections

11 © 2024 Nokia
 RAN node has context for two
RAN slice:

Domain Network Slice Stitching - RAN Slice #1 (i.e. S-NSSAI 1)

- RAN Slice #2 (i.e. S-NSSAI 2)

Example of End-to-End Network Slice Realization In context of RAN Slice #1,

UEs send IP flows with:
- 5QI 8
RAN node uses Mapping to 5QI - 5QI 6

▪ A single VLAN for each RAN slice ▪ DSCP 8 → 5QI 8 In context of RAN Slice #2,
▪ Multiple DSCPs for each IP flow ▪ DSCP 6 → 5QI 6 UEs send IP flows with:
- 5QI 9

gNB Router Router UPF

R1
Tunnel Blue
5QI 8 flow #1
UE1 RAN Slice 1
5QI 6 context VLAN 1, DSCP 8 Core Slice 1
Tunnel Tunnel
EP context
5QI 8 → DSCP 8 EP
5QI 6 → DSCP 6 flow #2
VLAN 1, DSCP 6
RAN Slice 1 → VLAN1
R2
RAN Slice 2 Tunnel Yellow
context flow #3 Core Slice 2
UE2 VLAN 3, DSCP 9 Tunnel Tunnel
context
5QI 9 EP EP
5QI 9 → DSCP 9

RAN Slice 2 → VLAN3

Example of how to stitch Transport Slice with RAN/Core Slices based on VLAN
12 © 2024 Nokia
Domain Network Slice
UE Slice
URSP (UE Route Selection Policy) is dynamic slice selection
technology enabling devices to utilize multiple slices in
parallel
▪ URSP contains multiple URSP rules providing a dynamic
matching of traffic to PDU sessions
▪ Identification of traffic based on Traffic Descriptor (TD)
▪ Mapping to existing PDU session or creation of a new one
based on Route Selection Descriptor (RSD)
▪ URSP requires 5G UE, 5G Core and Policy Control (SA)
▪ The encoding of URSP is defined in 3GPP TS 24.526
▪ UE receives the updated URSP rules and (re-)evaluates their
validities
▪ URSP is updated by PCF URSP Policy
Rule precedence: 1
▪ UE moved from EPC to 5GC
Traffic descriptor: Application identifier = OS ID + OS APP ID (Android OS ID + “ENTERPRISE”)
▪ Change of Allowed NSSAI Route precedence 1: sst: 1, sd: 1A2B3C, DNN selection “enterprise”
or Configured NSSAI Rule precedence: 100
▪ … Traffic descriptor: Match-all-type
Route precedence 1: sst:1 sd:123456, DNN selection “internet”

13 © 2024 Nokia
 Part II. Network Slicing Orchestration

14 © 2024 Nokia
Network and Service Orchestration
Key Standards and Open-Source Projects
Orchestration &
Control Layers Standards Activities Open-Source Projects

Service and Slice TMF

Management

MEF

ONAP Architecture

Linux Foundation Networking

Subcommittee
Domain-level
orchestration ZSM
Zero Touch

OSM
Network and
Service

Daylig
Open
Management
SA5
Domain-level
resource control NFV

Network resources
(physical/virtual)

15 © 2024 Nokia
 Slice Orchestration and Automation
TS 28.801 – Mgmt & Orch of Network Slicing for Next-Generation Network
Lifecycle of a Network Slice Instance
Preparation Instantiation, Configuration Run-time Decommissioning
and Activation
Design Pre-provision Instantiation/ Supervision
Activation Modification De-activation Termination
Configuration
Network environment Reporting
preparation

Preparation phase Instantiation, Configuration and Run-time phase Decommissioning phase

Activation phase
In the preparation phase the NSI In the run-time phase the NSI is capable of The decommissioning phase includes
During instantiation/configuration all
does not exist. The preparation traffic handling to support communication deactivation (taking the NSI out of active
resources shared/dedicated to the NSI have
phase includes the creation and services of certain type(s). The run-time duty) as well as the reclamation of
been created and are configured, i.e. to a
verification of network slice phase includes supervision/reporting (e.g., dedicated resources (e.g., termination or
state where the NSI is ready for operation.
template(s), the onboarding of for KPI monitoring), as well as activities re-use of network functions) and
The activation step includes any actions that
these, preparing the necessary related to modification. Modification configuration of shared/dependent
makes the NSI active, e.g. diverting traffic to it,
network environment which are could map to several workflows related to resources. After decommissioning the NSI
provisioning databases (if dedicated to the
used to support the lifecycle of NSIs runtime tasks, e.g., upgrade, does not exist anymore.
network slice, otherwise this takes place in the
and any other preparations that are reconfiguration, NSI scaling, changes of
preparation phase) etc. NS instantiation,
needed in the network. NSI capacity, changes of NSI topology,
configuration and activation can include
association and disassociation of network
instantiation, configuration and activation of
functions with NSI.
other shared and/or non-shared NF(s).

16 © 2024 Nokia
Slice Orchestration and Automation
TS 28.801 – Mgmt & Orch of Network Slicing for Next-Generation Network

CSMF CSMF translates communication requirements to network slices

End-to-end
Level
Network slice instance (NSI) monitoring, reporting and LCM
NSMF ▪ Slice level health monitoring and SLA assurance
▪ Slice life cycle management and closed-loop operations

Network slice subnet instance (NSSI) monitoring, reporting and LCM

NSSMF ▪ Alarm correlation and statistics aggregation on slice subnet level
▪ Network slice subnet (NSSI) life cycle management and
Domain provisioning according to slice profile
Level

NFMF Network function (NF) monitoring, reporting and configuration

NFMF
▪ Raw slicing alarm/counter collection on NF level
▪ NF slicing parameter configuration and provisioning

CSMF Communication Service Management Function

NSMF Network Slice Management Function
NSSMF Network Slice Subnet Management Function
Slice #1 NFMF Network Function Management Function

17 © 2024 Nokia
Slice Orchestration and Automation
GSMA NG. 116: Generic Network Slice Template

▪ GST (Generic Slice Template): A set of potential slice attributes that can
characterize a network slice. GST is generic and is not tied to any specific
network deployment

▪ NEST (NEtwork Slice Type): is a GST filled with values and is essential for
network slice providers to instantiate equivalent slices

[ GST Attributes ] [ NEST for URLLC ]

Availability, area of service, delay tolerance, deterministic communication,

downlink throughput per network slice, downlink maximum throughput per
UE, energy efficiency, group communication support, isolation level,
maximum supported packet size, mission critical support, maximum number
of PDU sessions, maximum number of UEs, performance monitoring,
performance prediction, positioning support, radio spectrum, Session and
Service Continuity (SSC) support, simultaneous use of the network slice, slice
quality of service, supported device velocity, synchronicity, UE density, uplink
throughput per network slice, uplink maximum throughput per UE, packet
delay budget, …

18 © 2024 Nokia
Slice Orchestration and Automation
3GPP TS 28.541 - 5G Network Resource Model (NRM)

S-NSSAI #a S-NSSAI #b S-NSSAI #c

Slice SLA Controller

S-NSSAI #a S-NSSAI #c
S-NSSAI #b

S-NSSAI #a S-NSSAI #b
S-NSSAI #c

Network slice instance X

NF5 NF6 Example of NSI X and Y

A service instance is realized by one or more network slice instances (NSIs),
Network slice subnet instance C

that in turn consists of network slice subnet instances (NSSIs) NF1 NF2 NF4 NF7 NF8

Network slice NF3 Network slice NF9

subnet instance A subnet instance B
Network slice instance Y

19 © 2024 Nokia
Slice Orchestration and Automation
3GPP TS 28.541 - 5G Network Resource Model (NRM)
Support
Attribute name
Qualifier
serviceProfileId M
pLMNInfoList M
maxNumberofUEs O
coverageArea
dLLatency
O
O Information model definitions
uLLatency
uEMobilityLevel
O
O for network slice NRM
networkSliceSharingIndicator O
sST M
availability O
delayTolerance O
dLDeterministicComm O
uLDeterministicComm O
dLThptPerSlice O
dLThptPerUE O
uLThptPerSlice O
uLThptPerUE O
dLMaxPktSize O
uLMaxPktSize O
maxNumberofPDUSessions O
Support
Attribute name
kPIMonitoring O Qualifier
userMgmtOpen O sliceProfileId M
v2XCommModels O pLMNInfoList M
termDensity O CNSliceSubnetProfile CM
activityFactor O RANSliceSubnetProfile CM
uESpeed O TopSliceSubnetProfile CM
jitter O
survivalTime O SliceProfile
radioSpectrum O
dLReliability O
uLReliability O
maxDLDataVolume O
maxULDataVolume O
nBIoT O
synchronicity O

ServiceProfile
20 © 2024 Nokia
Slice Orchestration and Automation
3GPP TS 28.541 - 5G Network Resource Model (NRM)

Attribute name Support Qualifier Attribute name Support Qualifier Attribute name Support Qualifier
dLLatency O coverageAreaTAList O maxNumberofUEs O
uLLatency O uEMobilityLevel O dLLatency O
maxNumberofUEs O resourceSharingLevel O uLLatency O
dLThptPerSliceSubnet O maxNumberofUEs O dLThptPerSliceSubnet O
dLThptPerUE O activityFactor O dLThptPerUE O
uLThptPerSliceSubnet O dLThptPerSliceSubnet O uLThptPerSliceSubnet O
uLThptPerUE O dLThptPerUE O uLThptPerUE O
dLMaxPktSize O uLThptPerSliceSubnet O maxNumberOfPDUSessions O
uLMaxPktSize O uLThptPerUE O coverageAreaTAList O
maxNumberOfPDUSessions O uESpeed O dLReliability O
nROperatingBands O dLReliability O uLReliability O
sliceSimultaneousUse O uLReliability O resourceSharingLevel O
delayTolerance O nROperatingBands O dLMaxPktSize O
energyEfficiency O dLLatency O uLMaxPktSize O
termDensity O uLLatency O sliceSimultaneousUse O
activityFactor O delayTolerance O delayTolerance O
coverageAreaTAList O sliceSimultaneousUse O energyEfficiency O
resourceSharingLevel O dLMaxPktSize O dLDeterministicComm O
uEMobilityLevel O uLMaxPktSize O uLDeterministicComm O
uESpeed O energyEfficiency O survivalTime O
dLReliability O termDensity O nssaaSupport O
uLReliability O survivalTime O n6Protection O
dLDeterministicComm O dLDeterministicComm O nonIPSupport O
uLDeterministicComm O uLDeterministicComm O
survivalTime O positioning O CNSliceSubnetProfile
positioning O synchronicity O
synchronicity O
nssaaSupport O RANSliceSubnetProfile
n6Protection O
nonIPSupport O

TopSliceSubnetProfile

21 © 2024 Nokia
 [1] “Applying Machine Learning to End-to-end Slice SLA Decomposition,” IEEE NetSoft
S4SI Workshop 2020

Slice Orchestration and Automation

[2] “Decomposing SLAs for Network Slicing,” IEEE Communications Letters 2020
[3] “Constraint-Aware Deep Reinforcement Learning for End-to-End Resource
Orchestration in Mobile Networks,” IEEE ICNP 2021
[4] “OnSlicing: online end-to-end network slicing with reinforcement learning,” ACM

E2E Slice Deployment & Runtime SLA Management CoNEXT 2021

[5] “Atlas: Automate Online Service Configuration in Network Slicing,” ACM CoNEXT 2022
[6] “RoNet: Toward Robust Neural Assisted Mobile Network Configuration,” IEEE ICC 2023

Digital Twins

E2E SLA Decomposition[1,2] DRL-based Runtime SLA Assurance[3,4] Digital Twins for E2E Orchestration[5,6]

Key E2E Slice Orchestration Intelligence

NSMF

NSSMF

Secure Slice Placement

5GC Auto-scaling Close Loop

22 © 2024 Nokia
5GC Orchestration Intelligence
End-to-End SLA to Domain SLA TMF 633 Service Catalog Mgmt
E2E Slice Orchestration - SLA Decomposition OSS/BSS
Customer Portal
TMF 641 Service Ordering Mgmt
TMF 638/639 Service Inventory
TMF 921 Intent Management API
…
NEST
E2E Service Orchestration

CSMF Service Profile

GSMA Generation from NSMF Slice Profile

GST/NEST OSS/BSS or as input
(TopSliceSubnetProfile)

RANSliceSubnetProfile TN Requirement* CNSliceSubnetProfile

(R)AN Domain TN Domain CN Domain

Orchestration rApp Orchestration Orchestration
O-Cloud
Non-RT RIC R-NSSMF T-NSSMF C-NSSMF
M&O

xApp Config Params Config Params Config Params

Near-RT RIC R-NFMF T-NFMF C-NFMF

RAN Transport Core

* IETF Yang model: https://datatracker.ietf.org/doc/draft-ietf-teas-ietf-network-slice-nbi-yang/)

23 © 2024 Nokia
 Concluding Remarks

24 © 2024 Nokia
Evolving Network Slicing Capabilities
Further Enhancement
3GPP Rel. 18 Network Slicing enhancements
▪ Better support for highly customized and localized service
▪ Dynamic network slices for temporary deployments
▪ Enhanced operator control for slice quota enforcement
▪ Enhanced operator control of when UEs can connect and use a particular network slice
▪ Service continuity in congestion and network maintenance
▪ Enhanced network slice management support

Considerations for further enhancements

▪ Huge number of slices, e.g., 2k
▪ Highly scalable and dynamic slice resource management
▪ Slice security, e.g., Network Slice Specific Authentication and Authorization (NSSAA)

25 © 2024 Nokia