end to end cognitive slicing - ieeesite.ieee.org/cscn-2018/files/2018/11/cognitiveslicing.pdf ·...
TRANSCRIPT
slicenet.eu
End to End Cognitive Slicing
1
IMEN GRIDA BEN YAHIA AND JOANNA BALCERZAK
ORANGE LABS NETWORKS, FRANCE
Outline Project ID card & Objectives
SliceNet Overview Information Model
Management Architecture
Stakeholders
SliceNet Use cases & Metrics Use case overview
QoE, QoS for slicing
Zoom on eHealth
Perspectives & conclusion
Call H2020-ICT-2016-2017
Topic: ICT-07-2017
Project type : RIA
Proposal number: 761913
Project duration: 36M
Budget:7.979.030 Euros
SLICENET: End-to-End Cognitive Network Slicing and Slice Management Framework in
Virtualized Multi-Domain, Multi-Tenant 5G Networks
Project ID Card
5G slice management and orchestration architecture for vertical businesses
Vertical-tailored control exposure of network slices
Defining cognitive techniques for Verticals request mapping, Slice resource optimization, QoE & QoS mapping
Setting transverse and common management building blocks across different vertical use cases
Project High level objectives
Outline
Project ID card & Objectives
SliceNet Overview Information Model
Management Architecture
Stakeholders
SliceNet Use cases & Metrics Use case overview
QoE, QoS for slicing
Zoom on eHealth
Perspectives & conclusion
SliceNet Slice Template (excerpt)
11
Class name Use case Smart City
Slice Type URLLC, eMBB, etc. mMTC
Network Performance Commited Bandwith per endpoint - DS
Commited Bandwith per endpoint – US
Total Slice Bandwidth – DS
Total Slice Bandwidth – US
50kbps
50kbps
the number of endpoints multiplied by the
committed bandwidth per endpoint(endpoints
communicate simultaneously)
Priority levels Latency – peak
Latency – mean
Jitter – peak
Jitter – mean
Packet loss - without
<1000ms
<500ms
300ms
100ms
<1%
Plug & Play feature Monitoring only
NFs configuration
QoS/QoE control
SDN forwarding
NFs lifecycle
Slice lifecycle
Monitoring only
Plug & Play view Service level
Slice level
NF level
Service level view
SliceNet Management & Orchestration
12
Co
gnit
ion
Su
b-p
lan
e
Orchestration Sub-plane
SS-O
NMR-O
Vertical
NFVI-PoP
P&P
Manager
NFVI-PoP NFVI-PoP NFVI-PoP
SliceNet Orchestration
multi-domain interactions
Service coordination
Slice orchestration
Resource orchestration
NFV-NS orchestration
Vertical Vertical
Mo
nit
ori
ng
Sub
-pla
ne
Infrastructure Sensors
…
Resource Monitoring
Raw resource, VNF,
PNF metrics/statistics
… …
Topology Monitoring
Network graph and UE
location
Traffic Monitoring
Traffic patterns and
characteristics
Slice Monitoring QoS Sensor
Slice end-to-end QoS metrics
Service Monitoring QoE Sensor
Service-level QoE
metrics/KPIs
Information Sub-plane
ST Catalogue
SD Catalogue
SI/NSI Inventory
NST Catalogue
Aggregation
Data Cleaning
Data Pre-processing
Data Storage
Analytics
Model Selection
Analysis Analysis
Analytic Engine
Model Training
Model Scoring
Policy Framework
Policy Recommender
Policy Repository
Policy Distribution
Model Inference Optimiser
Processing Engine
SliceNet Stakeholders
DSP & NSP Standalone Actors
DSP & NSP Combined Actors
16
Network Service Provider (NSP),
Digital Service Provider (DSP),
Digital Service Customer (DSC)
Outline
Project ID card & Objectives
SliceNet Overview Information Model
Management Architecture
Stakeholders
SliceNet Use cases & Metrics Use case overview
QoE, QoS for slicing
Zoom on eHealth
Perspectives & conclusion
Use case overview
eHealth UC
Enhanced Mobile
Broadband
Smart Grid UC
Ultra Reliable Low Latency
Communications
Smart City UC
Massive Machine Type
Communications
Aveiro, Portugal Cork, Ireland Alba Iulia, Romania
Smart Grid UC overview
Smart Grid UC
Ultra Reliable Low Latency
Communications
Aveiro, Portugal This UC relies on ultra-low latency (mesh) communications between the power grid sensors/actuators (IED - Intelligent Electronic Devices). It also requires high density of network communications coverage, as well as with high quality (in terms of ultra-low latency capacity). Based on these requirements, the most critical aspect of the UC is the RAN coverage.
E2E cognition requires well defined QoE and QoS to be granted
QoE definition QoS definition
Ref. by ITU –T P.10/G.100 Ref. by ITU –T E.800
QoE, QoS
QoE is therefore influenced by the delivered QoS (measurable with objective metrics) and the psychological factors influencing the perception of the user. The same QoS level might not guarantee the same QoE level for two different users.
Quality of Service is defined as the totality of characteristics of a telecommunications service that bear on its ability to satisfy stated and implied needs of the user of the service.
Quality of Experience has been defined as the degree of delight or annoyance of the user of an application or service.
Takeaway message
Towards E2E Cognitive Mgt. for slicing we need to revisit QoE, QoS for slicing
Identify the main management building blocks
On defining the slice template, data model for each use case
Endorse the work on data association, fusion, correlation and aggregation
Set up of integrated testbed allowing multi level orchestration and cognitive operations for the three project use cases and beyond.