5G Network Slicing Management for Challenged Network Scenarios Industry Indus try keyno keynote te 12th Work Worksho shop p on Challen Challenged ged Netwo Networks, rks, ACM ACM CHANTS, CHANTS, Octobe Octoberr 20, 2017 2017 Henning Sanneck E2E Mobile Network Solutions, Nokia Bell Labs Research, Munich, Germany with contri contributions butions from C. v. Hardenberg, Hardenberg, C. Mannweiler, Mannweiler, M. Naseer-ul-Islam, Naseer-ul-Islam, C. Sartori, Sartori, C. Schmelz et al.
Outline
5G
Conclusions
5G Network Slicing
5G RAN
5G Network Slicing Management
5G RAN Management
5G: just yet another “G” ?
Prof. Carle, TUM
Dr. Tsvetkov, TUM
Prof. Stadler, KTH
5G Cellular Networks New user demands with extremely diverse requirements
Devices 1.5 GB/day
Smart Factories 1 PB/day
Billions of sensors connected
Autonomous driving 1ms latency
5G: Diversity of Use Cases - The 4 Key Business Value Dimensions 10Gbps
360° video (free viewpoint)
Video
1Gbps
h 100Mbps t d i w 10Mbps d n a 1Mbps B
Autonomous vehicles
360° VR (hi-res) Remote training 360° video (lo-res)
SD Video streaming
100kbps Sensors
Virtual RAN
4k Video streaming
Things
VR/AR
Centralized RAN
Haptic VR Remote control vehicles
System Control
Cloud-assisted driving
Chatbots Electric grid control
10kbps Home Sensors
1kbps 10s
1s
100ms
Core Cloud
10ms
Latency
1ms
100us
10us
Edge Cloud
5G challenged network scenario: Industrial Internet / Industry 4.0 Resilient, secure low-latency communication Low / deterministic latency <1ms; 99.999% reliability
Wireline connections today
Manufacturing and process automation Resilient, secure low-latency comms
Inherent security by dedicated network slices
>90%
Single company network for all kinds of industrial applications
Critical comms
Intrusion detection
Overall costs for greenfield
2-5 times lower Public MNO slice
AR-enhanced maintenance
Break even for wireline replacement
1 year
# of sensors = Payback period
Reconfiguration cycle = Payback period
B u s i n e s s c a s e
Removing cost of cabling installation and maintenance
Less reconfiguration time Less production capacity overprovisioning
B e n e f i t s
Slicing
https://www.pexels.com/photo/wheat-bread-slices-166021/ http://tvtropes.org/pmwiki/pmwiki.php/Main/InventorOfTheMundane
5G Network Slicing | Optimized service delivery for heterogeneous use cases Multiple independent network instances on one physical network Slicing across radio, transport, core / edge and central clouds
Cloud scalability and efficiency
Autonomous driving
Self service portal Flexibility to meet diverse requirements
Industry Automotive Health
Health monitoring
Factory Full automation / self-organization
Different use case requirements different slice characteristics Latency Challenge
LOW HIGH
Mobile Broadband
HIGH SMF: Session Management Function
need
5G RAN: Requirements and Features
Services s t n e m e r i u q e R
s e r u t a e F N A R
UHD video
VR
Several 100 MHz spectrum New spectrum
Front haul split, HARQ Flexibility, New Radio (NR) Carrier Aggregation, NR-LTE Dual Connectivity
Ultra broadband
High Frequency deployments New spectrum
Support for Beamforming, Self Backhaul
VR
URLLC
Reduce latency to milliseconds
Drones
V2X
I4.0
Reliability
Instant response
At L1 or higher layers
Shorter TTI , RRC-INACTIVE, Zero latency Handover, SRB/DRB duplication, HARQ Flexibility
Multi-Cell Coordination: (CoMP), NR Multi-Connectivity, SRB/DRB duplication
5G RAN structure
= potential site for data center / aggregation
macro smallcells
edge cloud
tree
New radio New radio
small cells
RAN / Core network functions
RRHs
macro
MuLTEfire
chain
macro
4G WiFi
star
macro
Heterogenous environment: -Multi-RAT, multi-layer -Small cell cloud RAN -Physical and virtualized Network Functions -Multi-vendor
Core network functions
edge cloud
smallcells
ring
macro
central gateways
pre-aggregation
Distance and latency to radio access increases x100.000 small cells
x10.000 macro sites
x1.000 pre-aggregation sites
x100 aggregation sites
Very flexible, but also very complex RAN structure
x10 central gateways
5G RAN Management Addressing the challenges Ultra Dense Small Cells
Cloudified RAN & Core
Multiconnectivity (MC)
URLLC network service
Multi-service Network CP CP
Distribution
Centralization
Multi-RAT / layer
Low latency radio, edge cloud
Slices
MC-aware, management incl. aggregation of PM/FM data
NM data resolution; prognostic diagnosis; combined network resilience & selfhealing
Intra- / inter-slice management; Separation (& sharing) of knowledge; Embedded analytics
Flexible VNF re- location / re- configuration
Distributed decisions; Management hierarchy / aggregation
Combined management of physical / virtualized infrastructure
„Hybrid“ NM
5G RAN Management: Opportunity vs. Risk Opportunity: ubiquitous, unlimited connectivity for a wide range of services
Capacity
Risk: complexity of the network infrastructure (dense small cells, mixed physical / virtualized infrastructure)
Cost
prohibitive $
$$
„Unlimited“
Coverage „Ubiquitous“
Characteristic: Scale (# users, # applications) Manifestation: network usage data
viable
Network complexity
$
high
Characteristic: Scale (# cells, # (V)NFs) Manifestation: network operation data
Cognition: drive opportunity, limit risk by “mastering data” mastering complexity
Cognition & SelfOrganization
applied to infrastructure networks ?
Cellular macro network • Tightly planned, infrequent physical topology changes, automated operation • Single operator • Single vendor equipment per OAM domain 5G Cellular Heterogeneous Network • Some parts only coarsely planned, frequent virtual topology changes, highly automated operation • Multi-tenant (shared infra) • Multi-vendor per domain
Ad-hoc / mesh network • Uncoordinated deployment, frequent physical topology changes, autonomous operation • Only node operator • Open environment, standardized protocols between nodes
“Self-organization is a process where the organization (constraint, redundancy) of a system spontaneously increases, i.e., without this increase being controlled by the environment or an encompassing or otherwise external
5G RAN Management
verticals
IoT support
Cognitive Network Management System (multi-vendor, multi-tenant)
D2D
Virtualized Network Functions
Selfbackhauling
Optimization New radio New radio Ultra-dense small cells
Low latency
Low cost Multi-hop
High reliability Location information
Low power consumption
MuLTEfire
small cells macro
Troubleshooting / Healing Configuration
4G WiFi
Analytics
Policy
(Trained) telco-centric knowledge models & context network data
5G Network (Slicing) Management Analytics Engine
Knowledge Sharing / Isolation Verification
Communication Service Mgmt. (eMBB, mMTC, cMTC)
Network Preparation, LCM / (re-)configuration Slice Mgmt. Anomaly Detection Diagnosis Healing action Scaling Load balancing / traffic steering Scaling in/out, down/up Coverage and Capacity Optimization
Machine Learning
Mobility robustness (MRO)
Policy Engine
Objectives / Intent
Management Policies
Coordination
Anomaly Detection Diagnosis Healing
Anomaly Detection Diagnosis Healing
Neighbour relationship setup (ANR) (Big) data acquisition and distribution
Resource ID allocation (beam/cell ID/RS)
Network Function Chaining
OAM connectivity / interface setup
LCM / (re-)configuration, (re-)placement
workflow
MDT / location
Radio resources (beams, cells): PNFs
Cloud resources
Cognitive NM functions
5G Network Slicing Management: system architecture
Communication Service Management Network Slice Management InterSlice Mgmt.
NM
NFVO
OAM domain
NFV domain
EM
VNFM
VNF-App
PNF VNF-Plat
Communication Service Management has not been covered by 3GPP standards in the past • TMForum and ITIL provide (high-level) industry specifications
Network Slice management is “umbrella” functionality • Across radio / core / transport • Across infrastructure providers • VNFs and PNFs • Across vendors
Depth of control and entry levels of 3rd party 3rd party (e.g., vertical) MNO
Communication Service Mgmt.
Option 1 Web Service BSS
Network Slice Management
Network Slice Mgmt.
Option 2 Network Management ElementManager Manager Element
Network Orchestration VNFManager Manager VNF
Network Slice 1 Network Slice 2
BSS
Depth of control and entry levels of 3rd party 3rd party (e.g., vertical) Communication Service Mgmt.
MNO Web Service BSS
Network Slice Management
Network Management ElementManager Manager Element
Network Orchestration VNFManager Manager VNF
Network Slice Mgmt.
NM
Network Slice 2
Network Orchestration
Element Manager
Option 3
Network Slice 1
BSS
VNF Manager
Network Slice Lifecycle 3GPP 28.801 Lifecycle of a Network Slice Instance (NSI) Preparation
Instantiation, Configuration and Activation
Preprovision Network environment preparation
Design
Instantiation/ Configuration
Activation
NSI Automated Configuration
Decommissioning
Run-time
Supervision Reporting
Modificatio n
De-activation
NSI Automated Optimization / Healing
NSI Automated Re-Configuration Feedback for template optimization
Termination
Run-time optimization, reconfiguration Example of NF Overload detection CSMF
Trigger SLA modification negotiation Beyond SLA limits for the slice
NSI
Within SLA limits
NSI
review subnet provisioning
Modify subnet policy
NSMF
Beyond current subnet provisioning for the slice NSSI
NSSI
NSSI
NF Overload
within allowed limits for the slice
(e.g. increasing connected users
Each layer monitors and checks possible actions within its scope according to SLA. Otherwise „escalate“ to next higher level
Scaling or New Instance Instantiation
NSSMF
5G Cognitive NM function design Impacts of Multi-Tenancy and Multi-Service Cell-Specific Function Instance •
Differentiate input (performance metrics) for different slices
Slice-Specific Function Instance • Multiple instances per cell • Coordination between different function
•
Coordination among requirements of different slices
instances of same type operating on the same cell
•
Cognitive NM function policy update when slice is (de)activated
–
Parameters adjustable per UE
–
Parameters adjustable only at cell level Coordinator
Performance Metrics
CF1
Configuration
Network Slices Operating within a Cell
Performance Metrics
CF1-c CF1-b CF1-a
Configuration
Network Slices Operating within a Cell
5G Cognitive NM functions: example CCO / verification • Observed effect in real network data:
Cell B up-tilt 2° Cell A shows increased values of KPI “BLER in the HSUPA MAC layer” • Cause: Coverage and Capacity Optimization (CCO) algorithm
Cell C
triggered Cell C up-tilt Cell B forced to up-tilt due to being on the same RET module as Cell C
up-tilted (CCO) Shared RET
• Example case solved with hierarchical „network intelligence“:
CCO (tilt optimization) function plus verification function with wider network view
Cell B up-tilted (forced)
Tilt changes applied; Start of assessment process.
Hidden performance effect
Cell A Thur.
Fri.
Sat
Sun
-Increased HSUPA BLER -Neighbor of Cell B -Not neighbor of Cell C
5G Network Slicing Management for Challenged Network Scenarios Conclusions • 5G addresses some “challenged network” scenarios (factory, UAV, disaster response,
V2X) • 5G Network characteristics (ultra dense, cloudified, multi-service / -tenant) impose new RAN operability challenges • Functional: • per service- / tenant- instrumentation and dynamic operation (multiplicity of virtual network configurations) • data: higher resolution of measurements; new external sources / context • higher degree of autonomy in management • Architectural: • new building blocks slicing management, analytics & policy engine • higher degree of distribution, cooperation / coordination and abstraction Cognitive NM functions to master network data mastering network complexity
5G Network Slicing Management for Challenged Network Scenarios Conclusions • 5G Network Slicing Management • Different levels of control by tenant required (high-level vertical vs. MVNO) • Slice run-time optimization / healing slice-aware Cognitive Functions managing
Physical and Virtual NFs • Coordination across different slice’s requirements Enable the management of diverse service types in diverse network scenarios
Research challenges •
Dynamic slice instantiation and management (for unpredicted events)
•
Slice management knowledge sharing & isolation
•
Cognitive Function placement
•
5G URLLC management: instrumentation, prognostic diagnosis
Nokia Saving Lives
Nokia Saving Lives – https://networks.nokia.com/innovation/nokia-saving-lives Portable Control Center
UAV swarm
Portable LTE Ultra Compact Network (on pole, drone, balloon)
LTE
• • • • •
HD/IR/Thermal Cameras Loudspeaker / Microphone Gas Sensor Delivery System LTE data and control connection
LTE
Connectivity / PaCo Flight management Rescue Apps
RF BTS PaCo
Local NGO / International Crisis Management Organization
LTE
Data Storage
• Video streaming • 3D map creation • Object detection/People count • Flight management automation
(Mission Training, Storage, deployment) Local Operator (Frequency License)
Handsets
Nokia Operational Team/ Deployment