Post on 21-Jun-2018
transcript
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
Network Slicing Based 5G:
Mobility, Resource Management, and Challenges
Haijun Zhang April 29, 2017
University of Science and Technology Beijing
Email:zhanghaijun@ustb.edu.cn
Web: http://www.ece.ubc.ca/~haijunzhang/
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
Content
Background
System Architecture
Challenges
Mobility Management
Resource Management
1
2
3
4
5
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
1
Background
5G Drivers
5G Development
Guide
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
Mobile Internet and the Internet of Things (IoT) are two main drivers for future network slicing based 5G networks.• Exponential traffic and Massive data requirements• Ultra dense heterogeneous networks• Cloudization and virtualization• "Mobile" out, wireless "connectivity" in• New business model• Energy challenge• Nobody can "make" more spectrum…
5G Drivers
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
5G Development
A concept for future mobile and wireless communications system that supports the connected information society.• An end-to-end ecosystem• Network as a service• Fully mobile and connected society • Consistent user experience• Sustainable business models
1000xhigher mobile data volumes
10x – 100xhigher number of connected devices
10x – 100xtypical end-user data rates
5xlower latency
10xlonger battery life for low-power devices
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
2
System Architecture
System architecture
Slicing management
Guide
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
WiFi
Mesh
MEC
Data Center
Content Server
Management Server
Edge Cloud
Core Cloud
Network Slicing
Bac
khau
lF
ront
hau
l
Access Unit Access UnitAccess Unit
uRLLC SliceIoT Slice
NFV&SDN
Control plane
Data plane
eMBB Slice
Control plane
Data plane
Control Unit
Data UnitCache
Cloud Unit
Cloud Unit
Control Unit
Data Unit
Control Unit
Application V2X
Data UnitV2X
Cloud Unit
Centralization
Distribution
Fro
nth
aul
Fro
nth
aul
Bac
khau
l
Bac
khau
l
Macro BS
System Architecture
➢ Substantial change of
5G network architecture
➢ Perfect combination of
NFV & SDWN
➢ Heterogeneous ultra
dense small cell
networks
Network slicing based 5G system architecture
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
An end-to-end network slice is a specific collection of network functions and resource allocation modules isolated from other network slices.
uRLLC Slice
eMBB Slice
IoT Slice
NFV&SDNVir
tuali
zed
Reso
urc
es
Ph
ysi
cal
Resou
rces
Compute
Storage
Network
Compute
Storage
Network
Compute
Storage
Network
Compute
Storage
Network
Access
NetworkEdge Cloud
Core
Cloud
Netw
ork
S
lices
Application
Control plane
User plane
Dedicated Physical
Infrastructures
SDN
ControllerVNFM
Management & Orchestration
VIM
Network Slicing Management
VIM: Virtualized Infrastructure Management
VNFM: VNF Management
➢ enhanced mobile
broadband (eMBB)
slice
➢ ultra-reliable and low-
latency communication
(uRLLC) slice
➢ Internet of things (IoT)
slice
Slicing Management
Network slicing management with SDN & NFV
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
3
Mobility Management
Necessity
Mobility management
Guide
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
Network slicing based 5G systems will still face mobility management challenges caused by the potentially ultra high density of 5G networks combined with highmobility and high density of end devices.
Necessity
➢ the simple and single RAT
handover cases have evolved
to managing complex, multi-
RAT mobility scenarios
➢ the integrated control
functions can reduce control
signaling even for massive
distribute network nodes
➢ different network slices have
different requirements for
mobility, latency and reliability
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
Mobility Management
UETarget
Access UnitCore Cloud
Source
Edge Cloud
Target
Edge Cloud
Source
Access Unit
Measurement ControlArea Restriction Information Provided
Packet Data Packet Data
UL Allocation
Measurement Reports Measurement Reports
HO Decision
HO Request
Admission Control
HO Request ACK
HO Request ACK
Handover CommandHandover Command
Detach from Source
cell & Synchronize to
Target cell
Deliver buffered & transit
packets to target cell
Data forwarding
Data forwarding
Buffer packets
from source cell
Synchronization
DL Allocation
UL Allocation + TA for UE
SN state transfer
Handover Confirm Handover Confirm
Packet Data Packet Data Packet Data
Path Switch Request User Plane Update
Request
Switch DL pathEnd Marker
Packet Data Packet Data
End Marker
End MarkerEnd Marker
User Plane Update
ResponsePath Switch Response
Ack
UE Context Release
Release
Resources
Handover
Preparation
Handover
Execution
Handover
Completion
L3 Signaling L1/L2 SignalingUser Data
➢ Mobile devices register their locations when they first connect to the network, and then report their location information to the network periodically.
➢ The HSS will be distributed into the edge cloud, making them closer to end devices to shorten registration delays and reduce backhaul burdens.
➢ 5G networks will aggregate multiple heterogeneous RATs.
➢ Multi-RAT coordination is needed for different RATs to share location information of their mobile devices for achieving unified multi-RAT access and seamless mobility in 5G networks.
Location registration Handover management
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
SDWN(5G)
4
Resource Management
System modelling
Problem formulation
Simulation results and
discussions
Guide
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
System Modelling
➢ A suburban environment is considered with small cells randomly distributed in the macrocell coverage area.
➢ Uplink transmission resource allocation is developed including joint subchannel and power allocation.
➢ Small cells receive two kinds of interference: • cross-tier interference from the microcell;• co-tier interference from neighboring small cells.
Scenario assumption
➢ Coverage radius: macrocell, 500 m; small cells: 10 m.➢ Carrier frequency: 2 GHz.➢ Channel bandwidth: 10 MHz.➢ Minimum inter-small-cell distance: 20 m.➢ Macrocell user (requesting IoT slicing services) number: 50.➢ Each small cell user (requesting eMBB and uRLLC slicing services)
number: 2 or 4.
System parameters
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
The uplink resource allocation problem is developed as the maximization of uplink capacity on each subchannel for small cells considering the following constraints:
Problem Formulation
, , , ,
, , , ,, f 1 1 1
maxF
f u n f u n
F U NF
f u n f u na p u n
a C
, , , , max1
, , max
, , , ,1
, , , , , ,1 1
, ,1
, ,
s.t. 1 : , ,
2 : 0 , , ,
3 : , ,
4 : ,
5 : 1, ,
6 : {0,1}, , ,
NF
f u n f u nn
F
f u nN
F
f u n f u n u eMBBnF U
F FM th
f u n f u n f u n nf uU
f u nu
f u n
C a p p f u
C p p f u n
C a C R k u D
C a p g I n
C a f n
C a f u n
Objective function:
Constraint conditions:
a non-convex discrete objective function
➢ a convex continuous function
(by relaxing the binary subchannel allocation indictors)
➢ K×N sub-problems
(Lagrangian dual decomposition method)
• the optimal power allocation
(KKT conditions)
• the optimal subchannel allocation
(sub-gradient method)
• maximum transmit power
• minimum data rate requirement
• threshold of total interference power
• subchannel scheduling policy
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
Simulation Results and Discussions
eMBB uRLLC IoT
➢ The eMBB slice capacity rises nearly linearly with the density of small cells and increases
slightly with the number of users per small cell.
➢ The total capacity of uRLLC slice increases with the number of small cells, but the capacity
of uRLLC slice is 20 times less than that of eMBB slice.
➢ The total capacity of the IoT slice decreases with the number of small cells, due to the
increasing cross-tier and co-tier interference.
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
5
Challenges
Network reconstruction
Slicing management
Cooperation with other
5G technologies
Guide
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
• Both RAN and CN need reconstruction to support end-to-end network slicing.
• Especially in UDHSNs, not only the cooperation of macrocells and small cells should be designed to meet the customized slicing demands.
• The cooperation of multiple RATs should be considered to provide seamless mobility and high transmission throughput.
Network reconstruction
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
• Supporting customized configuration of resources, management models, system
parameters for various use cases in an isolated or abstract way.
• Service providers and operators have started developing industrial solutions for
network slicing.
• To create, activate, maintain and deactivate network slicing at the service level.
• To adjust load balance, charging policies, security and QoS at the network level.
• To abstract or isolate virtualized network resources; inter-slice and intra-slice
resource sharing.
Slicing management
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
• Network slicing needs to coexist and cooperate with traditional wireless network
technologies (such as broadband transmission, mobile cloud engineering (MCE),
SDN and NFV) evolved from LTE/LTE-A systems.
• However, there is still no proper approach to integrate network slicing with C-RAN,
SDN and NFV.
• The virtualized cloud of access networks and CN have the advantages of physical
resource pooling, distribution of software architectures, centralization of management.
Technology Cooperation
Air Interface
• Higher frequency bands
• NOMA
• massive MIMO
• mmWaves
• CR
Network Deployment
• Ultra-dense
• Heterogeneous
• Small cell
• D2D
• V2X
Virtualization
• User/Control plane spilt
• Cloudization
• C-RAN
• SDN/NFV
• MCE
http://www.ece.ubc.ca/~haijunzhang/ dr.haijun.zhang@ieee.org
Thanks!