Cognitive Network Control for 5G · •Application layer protocol should be interfaced with the...
Transcript of Cognitive Network Control for 5G · •Application layer protocol should be interfaced with the...
1 | Copyright © 2018 Tata Consultancy Services Limited
Cognitive Network Control for 5G
Hemant Kumar RathSenior Scientist,
TCS Research and Innovation,Bangalore/Bhubaneswar
India
Sept 2018
‹#›
Introduction - 5G Networks
5G Use Cases
Managing the Network
Cognitive Control
Middle Layer based solution
Conclusion
Copyright © 2018 Tata Consultancy Services Limited
Use Cases – Vertical Industries Categorization
• Industry process automation, automated production line, supplychain management, warehouse managementAutomated Factories
• Assisted driving, autonomous driving, in-vehicle mediaAutomated Transport
• Robotics Surgery, remote health monitoringHealthcare
• Smart grid, smart meterEnergy
• Real-time inspectionInsurance
• Remote education, skill developmentEducation
3
Copyright © 2018 Tata Consultancy Services Limited
Use Cases – Vertical Industries Transformation
Robotic Surgery
Modern Warehouse
Modern Industry
Rescue Operation
Remote Machine Maintenance
4
Copyright © 2018 Tata Consultancy Services Limited
Use Cases - Categorization
URLLC
mMTC
eMMB
5
Copyright © 2018 Tata Consultancy Services Limited
Understanding the Challenges
• Guaranteed communication with mobility support
• Link, coverage and congestion prediction and handoff between stations
• Moderate to huge data volume - 360 degree stereo cameras !
• Moderate to low packet error rate - 10-6 to 10-10 !
• Extremely low to moderate latency requirement: 10 msec for applications, 200-500 µsecs for machine control
• Cost effective, reliable, stable, scalable and vendor-independent solutions
Transformation Challenges
• Quality control, data analytics, coordination, Inventory control support
• Minimal configuration, per-flow provisioning, parameterization of QoS
• Application layer prospective - technology and protocol agnostic
• Solution for Greenfield as well as existing use cases
Operational Challenges
6
Copyright © 2018 Tata Consultancy Services Limited
5G, the Answer…
Network Speed
• 10 GB/s – 100 times faster than 4G
Latency
• 1 ms to support Augmented Reality and Tactile Internet
Density
• Very dense, 100000+/km2
Cost & Energy
• Low
IoT support
• Must
Mobility
• beyond 350 km/h
An agreement to set of new requirements for wireless
comm. systems that mature beyond 2020
7
Copyright © 2018 Tata Consultancy Services Limited
Five G’s of 5G
•10 Gbps or more
Gigabit Speed
•Above 6 GHz
Gigahertz frequencies
•Rapidly adapt to a broader range of requirements and demands
Greater Flexibility
•M2M opportunities for billions of things
Gizmos and Gadgets
•Global race for leadership – device manufacturers, service creators, app developers and operators
Global Competitions
8
Copyright © 2018 Tata Consultancy Services Limited
5G Realization!
Newer Frequency
bands are to be considered
More and more cells required
Mix technology
will rule
9
Copyright © 2018 Tata Consultancy Services Limited
Network Topologies in 5G - what is desired
• Forwarding plane – Simplifications, sustainability, performance, unified control policy, guaranteed scheduling
• Centralized control plane – latency reduction
• Three clouds – Access, Control and Forwarding clouds
Decoupling of Control and Forwarding plane in the GW
10
Copyright © 2018 Tata Consultancy Services Limited
Network Topologies in 5G - what is desired
Divide the control plane with more and more logical units
•Optimized process logic
•Signaling delay can be reduced
•Diverse networks can be realized and easy for standardization
State separated core
•State information (mobility) and processing logic plane can be decoupled
•State information can be centrally available
•Dynamic load management can be made possible
Control Function Reconstructions
Connection parameters will be diverse
•Terminal capability, location and mobility history, request type, service feature, etc.
4G Mobility management protocols GTP and PMIP should go
•Unified mobility management protocols needed
•Mobility as a service, location aware services, etc
• Improved vertical and horizontal handover
New Connections and Mobility Management
11
Copyright © 2018 Tata Consultancy Services Limited
5G Network Characteristics (1/2)
• Inter-cell coordination
• Improved resource usage
• Different service provisioning
• Different network topographies
Access Plane
• Centralized control functioning
• Globally resource scheduling
• On-demand orchestration
• New Service exposure layer
Control Plane
• Simple gateway functions
• Distributed deployment
• Low latency and high data rateForward Plane
12
Copyright © 2018 Tata Consultancy Services Limited
5G Network Characteristics (2/2)
• Mesh and ad-hoc networks
• Radio resource sharing
• Customized network and services
• Awareness and treatment
• New network technologies
Access Plane
• Control function reconstruction
• Mobility Management
• Network Capacity Exposure
• Value added services
Control Plane
• Gateway C/U Split
• Mobile edge contents and computing
• End-to-end requirements Forward Plane
Multi-RAT Convergence
On-demand Networking
13
Copyright © 2018 Tata Consultancy Services Limited
Network Topologies in 5G – Key Points
On-Demand Networking
• Network orchestrator plays the major role
• NFV/SDN brings vritualization
Multi-RAT Cooperation
• Intelligent access control and management
• Wireless resource management and Dynamic load management
• Protocol and signaling optimization
• Multi-node and connections
• Spectrum sharing can be thought of
14
Copyright © 2018 Tata Consultancy Services Limited
Network Architecture for 5G
Service layer architecture for 5G
• Micro service based architecture for auto scaling with access network focus• Multi-operator and
multi-technology support: Hybrid network support –WiFi, LTE/LTE-A, LAA…
• Service based slicing. vs. application based slicing vs. flow based slicing
Vertical business services supt. – apps
with comm. c/s
• Application aware network slicing, virtualization, cloud management, prediction and control
• Viz: Automotive Vehicle: Driving Ast(URLLC), Ent. (eMBB), Sensors (mMTC)
5G Overall Architecture (5GPPP)
15
Copyright © 2018 Tata Consultancy Services Limited
SDNization – Possible Solution
• Adaptive applications and networks
• Should be able to use for existing and green-field solutions
• Dual connectivity – New Radio can be realized
Application layer specific solution to manage both the application and lower
layer protocols
• Independent application layer modules to be defined which are to be used as the core for a new service
• Service or application agnostic modules vs. application dependent modules
• Brings programmability and separation of network with slices – same hardware
• Slicing is possible
SDN vision for control and data extended to
applications
• Application should be flexible to run in different networks & devices
• Networks should be flexible, scalable and easily configurable
• No physically separate network for different application/services
SDNization requires adaptive applications and
networking
16
Copyright © 2018 Tata Consultancy Services Limited
Translating the SDNization (1/2)
• QoS guarantee, compute and communication issues
Understanding the applications/use cases
• Kind of network, protocols, controllability of the network and protocols
• Real-time behavior of the network and protocols
• Capability of the underlying elements
Understanding the network
• Based on the application requirement and network behavior
• Can be real-time or off-line
Taking a decision
• Fine tuning the application, controlling the protocols and provisioning the network
• Deciding the optimal deployment decision based on optimization
Deploying the decision
17
Copyright © 2018 Tata Consultancy Services Limited
Translating the SDNization (2/2)
• Application parameters can be decided based on the service definition
• Real-time application parameters can be tuned based on the network data
• Device data also plays a role
• Telematic and sensed + synthetic data have different QoS requirements
• Application layer protocol should be interfaced with the cross layer data
• Data collection/feedback – is a key
Network–aware Applications
• Network provisioning can be performed based on the application parameters
• Provisioning includes scheduling, path selection, routing, congestion control, power control, interference management
• Modules to be written which can be interfaced with the applications and appropriate commands can be generated to appropriate nodes/links/path
Application-aware Networking
18
Copyright © 2018 Tata Consultancy Services Limited
Orchestration and Management
• Should be capable of end-to-end orchestration
Fully automated and real-time orchestrator is required
• Multi-domain service orchestration: spanning across admin, operator and user (!!) domains
• Automated analytics and end-to-end FCAPS (fault, configuration, accounting, performance and security) support
• Heterogeneous network and application support
• Self-aware, self-healing, self-management….
Key Functionalities
• 5GPPP – Measure, Analyze, Policy and Execute (MAPE)
• ETSI – NFV Management and Orchestration (MANO)
• MEF – Lifecycle Service Orchestrator (LSO)
• TMF – Zero-touch Orchestration, Operation and Management (ZOOM)
Possible architectures today
19
Copyright © 2018 Tata Consultancy Services Limited
Cognitive Network Control – Overall Architecture
Decide
Analyze
API Layer
Applications
Application
Sensing
Respond
RSSI, Configuration, congestion, route, link
state, port & flow parameters
Network, Protocol
and Application
state - Logs, Real-
time params
Current State: Features,
synchronized params, QoE
params
Protocol
parameters,
desired state,
iterated
parameters
Configuration, Slicing,
Virtualization, Power, Route ,
control…
KPIs &
App
params
L1-L4
params
for cross
layer opt
Sense
ApplicationNetwork &
Protocols
Device & Network
Layer
SDN NFV Legacy
SADR Framework
• Sensing of Application and
Networks
• Sensing can be triggered or
automatic
Machine Learning
Classification
Feature Extraction
Model Tuning
SDNization can be
achieved through a
cognitive control
framework
20
Copyright © 2018 Tata Consultancy Services Limited
Cognitive Network Control – Sensing & Analysis
Sensing
• Sensing of device, network and application data
• Probe agents can be used along with the controller for sensing
• Type of sensing
• Active, Passive/Indirect Sensing
• Device & Services
• Browsing habit, location, CDR
• Edge & Core
• KPIs, flow-level, service-level, user-level
• DC & Cloud
• Application specific, Security and privacy
Analysis
• Derived Analysis
• Root cause analysis, understanding the fault
• Predictive & Proactive Analysis
• Expected problem and main causes
• End-to-end big data analysis
• Event pattern and relationship
• Market analysis – possible !!!
• Prescriptive Analysis
• Recommendation for specific action
• Estimation of future action plan
Machine Learning
Classification
Feature Extraction
Model Tuning
Note: Regulatory issues play a role in
collecting the data; all the data may
not be available
21
Copyright © 2018 Tata Consultancy Services Limited
Cognitive Network Control - Decision
Output
Input
Emulated Nodes
Simulated Network Topology
External Real Network
Applications
Model
Classification
Algorithms/Models
Emulation
Model Validation
To Respond
Sensed &
Analyzed data
Training
Decision – Rule based or Cognitive
• Optimal vs. sub-optimal solution
• Use case analysis is required – use of extended emulation module to recreate the scenario and to form the rule or models.
• These rules or models are to be trained further and appropriate emulation based testing is required for validation
• Any decision to be deployed (as a part of respond framework) has to be validated in the emulator
• Along with the observed data (sensed data), synthetic data generated by the emulation are to be used for modeling, training and decision making
22
Copyright © 2018 Tata Consultancy Services Limited
• Real-time sensing of network/application/protocols
• Cognitive network prediction & provisioning
• SDNization - SDNized APIs in the south bound
• APIs support for OpenStack and other Cloud orchestrators
Key Features
• Middle layer based solution
• Auto sensing, analyzing, prediction & configuration
• Cognitive control of network and protocol
• Vendor and technology agnostic solution
• Heterogeneous network support
• Emulator based decision making
Novelties
Cognitive Network Control – Our Solution
23
Copyright © 2018 Tata Consultancy Services Limited
Cognitive Smart Engine Operation – 5GPPP
Monitor the Traffic (eMBB, mMTC, URLLC), Mobility, Radio + KPIs
Analyze the Achieved/Projected performances, Real-time learning
Plan for new RRM Algos, D2D context, Spectrum usage, dynamic slicing
Execute with virtual or physical N/W slicing & orch., SDN & NFV support, Mode selection
24
Copyright © 2018 Tata Consultancy Services Limited
Cognitive Controller – Solution Architecture
Application
Respond
Analyze
SenseMiddle
Layer
Decision
Network Control Controller
Network
Control AgentNetwork
Devices
• Cloud based application
• SDN based network softwarization to realize network slicing
• Support for HetNets
• Rule based decision making process
• Learning modules – network, usage, QoS, applications
Middle layer based solution
• Applications with all possible traffic type
• Applications with single traffic type – mainly video application, smart city applications (smart meter!)
• Vertical industries – Healthcare, factories, multimedia, automative, energy
Applications & use cases
Emulation
25
Copyright © 2018 Tata Consultancy Services Limited
What Can be Achieved ? – Multi-tenancy, 5GPPP
26
Copyright © 2018 Tata Consultancy Services Limited
Multi-tenancy through Cognitive Control, 5GPPP
27
Copyright © 2018 Tata Consultancy Services Limited
References
1. View on 5G Architecture (5G PPP White Paper), Ver. 2, Jul 2018
2. 5G-PPP, Living Document on 5G PPP use cases and performance evaluation models, 2014
3. Study on Architecture for Next Generation System – Sept 2016
4. 5G – Personal Mobile Internet Beyond What Cellular Did to Telephony, G Fettweis et al., IEEE Communication
Magazine, Feb 2014
5. A Study on 5G V2X Deployment (5G PPP White Paper), Ver .1, Feb 2018
6. 5G Vision, 5G PPP, 2012
7. NGMN Alliance 5G White Paper, Feb 2015
8. 5G Service-Guaranteed Network Slicing White Paper, Feb 2017
9. 5G Services and Use Cases, 5G Americas White Paper, Nov 2017
10. Cognitive Network Management for 5G, Robert Mullins et al., Waterford Institute of Technology, Mar 2009
11. 3GPP TS 38.413: NG Application Protocol (NGAP).
12. 3GPP TR 21.866: Study on Energy Efficiency Aspects of 3GPP Standards.
13. 3GPP TS 23.214: Architecture enhancements for control and user plane separation of EPC nodes.
14. 3GPP TS 23.501: System Architecture for the 5G Systems
15. 5G Network Architecture, A High-Level Perspective, Huawei, 2016
16. Draft Recommendation Y.IMT2020, reqts, “Requirements of IMT 2020 network”
Disclaimer: Some of the figures used in this presentation are taken from most of the above references and these figures are
used only for academic purpose and illustration of the concepts.28
29 | Copyright © 2018 Tata Consultancy Services Limited
Thank you…Thank you…
Email: [email protected]: [email protected]