An Intelligent SDN

Framework for
5G Heterogeneous
Networks
Guided By:
Mr. Premkumar C.V
By
NIVEA KESAV
SEMESTER 2
ROLL NO : 10
WIRELESS TECHNOLOGY
SOE ,CUSAT
 OVERVIEW
Major challenge in 5G -improve system capacity and meet dynamic service
demands
Promising solution Heterogeneous networks (HetNets)
Proposing a software defined network based intelligent model
How to efficiently manage the heterogeneous infrastructure and resources
Reviewing the latest SDN standards and discussing the possible extensions
Advantages of SDN in meeting the dynamic nature of services and
requirements in 5G HetNets
Analyzing variety of schemes to improve
- traffic control
- subscriber management
- resource allocation
 WHY 5G ????

Very High speed, high capacity, and low cost per bit
Supports interactive multimedia, voice, video, Internet, and
other broadband services
Offers Global access and service portability
Providing large broadcasting capacity which supporting
almost 65,000 connections at a time
More applications combined with artificial intelligent (AI)
Use remote management that user can get better and fast
solution
Based on All IP Platform
Measurement of traffic in order to provide Quality of Service
CHALENGES FACED IN 5G
Efficiently meet the increasing demand for network
capacity while spectrum resource remains scarce

Handling the complex context of operations

characterized by a tenfold increase in traffic

Guaranteed quality of service (QoS)

One possible solution

-HetNets
HETEROGENEOUS NETWORK
Indicates the use of multiple types of access technologies
Co existence of macrocells, picocells or femtocells
Device to Device (D2D) communications
Problems to be solved in heterogeneous wireless networks such as:
-Interoperability of technology
- Handover
- Mobility
- Quality of Service
- Interference between RATs
Proposed Solution SDN framework
SOFTWARE DEFINED NETWORK
Decoupling of the control plane and data plane

Logically centralised control

Suitable for the dynamic nature

Characteristic features
- Directly programmable
- Agile
- Centrally managed
- Programmatically configured
ARCHITECTURAL COMPONENTS OF SDN
SDN APPLICATION

SDN CONTROLLER

SDN NORTH BOUND

INTERFACE

SDN SOUTH BOUND

INTERFACE

SDN DATA PLANE

EXISTING SDN STANDARDS

Open networking forum (ONF)

- using open flow standards
Internet Engineering Task Force (IETF)
The International Telecommunications Union
Telecommunication Standardization Sector (ITU-T)
The European Telecommunications Standards Institute
(ETSI)
The China Communications Standards Association (CCSA)
NECESSARY STANDARD EXTENSIONS FOR
FUTURE 5G NETWORKS
Requires necessary extensions to current standards

The switching of multi-homing user equipment between HetNets easily

managed without sudden loss of connectivity

IEEE 802.11f,802.11k, 802.11r for the homogeneous handover situations

IETF mobile IP (MIP) for handoff between the networks of different service
providers
ADVANTAGES
SYSTEM MODEL OF INTELLIGENT SDN
Inheriting the system architecture evolution of 4G LTE
 Core network (CN) - Evolved packet core (EPC) is comprised of
-serving gateway (S-GW)
-mobility management entity (MME)
-packet data network gateway (P-GW)

The radio access network (RAN) -Evolved-universal terrestrial radio access

network
(E-UTRAN), is made up of
-macrocell BSs(also known as evolved NodeBs (eNBs))
-pico, femto and relay nodes
 CN supports two types of operations:
-request for network association from a UE
-the networks response
S-Random
memory
CN mainly include three parameters:
-utility of the CPU
fup -Uplink
c-cpu memory -the size of the storage
Frequency

-data bandwidth

V = (c, , , s, nu, p)

p- transmitted fdn -Downlink

power Frequency

nu-number of
users attached
to the nodes
APPLICATION SCENARIOS
Scenario 1Multimedia Services:
Comparing resource consumption of three types of multimedia services:
-live video conferencing
-live TV streaming
-HTTP adaptive streaming (HAS)
 For video conferencing service:
-network should provide a multi-point to multipoint connection
-low latency and delay variation
-indicates fast data forwarding and stable bandwidth
-data storage and processing may not be required as much
-the demand for uplink and downlink data transmission should be much more
urgent than for the CPU and storage.
For live TV streaming
-one-point to multi-point service
-downlink bandwidth consumption is the dominant issue
-requires limited CPU and storage resources
For HTTP adaptive streaming
- encoder is involved in generating video slices of dynamic bit rates
- all four aspects of resource requirement are more intense
than the foregoing cases
Scenario 2Network Load Variation:
Network load usually varies dynamically over time
Analysing daily network load fluctuation for business and residential areas

Network load variation: a) daily network load variation; b) emergency and hotspot events

For a business area -the network load becomes relatively high in the daytime but low at
night
For a residential area- Opposite behaviour
 Network must be aware of and adaptive to traffic
fluctuations in both time and geographic dimensions

Emergency or a hotspot event - sudden burst in network

load

Network should be intelligent enough to handle this

situation

Useful technology -eMBMS technology

Multimedia data can be received by a device in Unicast

or Broadcast mode
 Need for
Inference dynamic
adaptation
DIFFERENT INTELLIGENT SCHEMES

Traffic Behaviour Prediction:

Increasing traffic demand of users results in need for traffic
prediction
Accurate prediction of traffic requires tracking fluctuations of data
flows dynamically in real time
Two main principles for traffic prediction

real-time methods machine learning methods

Real-time method :
System takes the latest obtained data profile as the one that will hold
into the future
Only in the case of slight variation from time to time we can achieve
good performance

Machine learning methods :

Use of back propagation algorithm
Multi layered feed-forward networks
For traffic prediction, a hierarchy of three layers is sufficient
Mainly four mechanisms involved
Load Balancing:

It is necessary to allocate time and frequency resources efficiently due

to insufficient spectrum consumption

Each macro cell is provisioned to possess an equal amount of resources

Problem occurs if users are in motion

Homogeneous quality of service cannot be ensured over the entire

network

An efficient load balancing algorithm that makes full use of resources

globally is necessary
 Load balancing between macro cells and femto cells
-crucial issue to deal with
Feasible and dynamic solution to inter cell load balancing
- Single server processor sharing queue
In a single server PS queue the capacity C of the server is
equally shared between the customers in system
If there are n customers in system each receives service at the
rate C/n
Customers dont have to wait at all; the service starts
immediately upon arrival
 PS queue is an idealized model

For the round robin (RR) discipline the customers are served in turn,
each for a small time slice

For document or file transfer, these are divided in small packets served
in turn

User Density Prediction:

Density distribution of users in a cellular network provides valuable
information that influences location based services and applications

With accurate user density - network is able to allocate resources more

wisely
 User density can be predicted with power allocation distribution
- high power sensed in one area usually indicates
- high user density in that area

Concept of SFFR

Hard frequency reuse ????

FFR - Users are classified into cell-edge users and cell-center users
 Analysis of fractional frequency reuse at cell edges
- frequency reuse bandwidth

inter-cell interference SIR

Case when reuse factor is 1 or 2

Case when reuse factor is 3

Case when reuse factor exceeds 3

 14
12
10
8

Cell edge SIR

6
4

(dB)
2
0
-2
-4
-6
1 2 3 4 5 6 7
reuse factor

SIR of cell edge changed with reuse factor

Cell edge channel capacity

20

15
(Mbps)

10

0
1 2 3 4 5 6 7
reuse factor

Cell edge channel capacity changed with reuse factor

 In soft frequency reuse scheme inter-cell interference at the cell edge is
greatly reduced through increasing frequency reuse , which improves the
channel capacity
MS is in the center of cell, the received power of the wanted user signal is
strong, and the interferences from other cells are weak
In the cell center, all the sub-carriers can be used to achieve high data rate
communication
SFFR characterized by
- frequency reuse factor 1 in the central region of a cell
- frequency reuse factor greater than 1 at cell edge region
Dynamic allocation of mobile users to cell centers and cell edges is done
using intelligent SDN mechanisms
 Assume having reuse factor 3 is used in cell edge and reuse factor 1 is used in cell
center
Radio Resource Allocation:

In HetNets, radio resource management is a key to jointly managing

resources from multiple access networks.

SDN controller aims to jointly manage resources among the networks to

maximize network utility

Traffic is tidal in nature

This makes RRM and performance analysis more complex and difficult
 Proposed solution : Inter-Tier Spectrum Sharing

Several strategies are chosen to obtain a leader and followers

The MeNB is the leader and PeNBs the followers

The MeNB shall impose a price on the PeNBs for shared frequency
bandwidth

The more bandwidth the PeNB demands, the higher expenditure PeNB
has to pay
CONCLUSION

SDN controller takes the responsibility of managing the

infrastructure and optimizing resources

Factors like traffic control, load balancing, density

prediction and resource allocation were analysed

Number of smart schemes to overcome these challenges

REFERENCES
[1] Songlin Sun, Liang Gong, Bo Rong, Kejie Lu, An Intelligent SDN Framework for 5G
Heterogeneous Networks IEEE Communications Magazine, vol.53, no. 11, Nov. 2015, pp.
142147.
[2] Shui Yu, Yong Xiang, Keshav Sood, Software-Defined Wireless Networking
Opportunities and Challenges for Internet-of-Things: A Review IEEE Internet Of Things
Journal, vol.3, no. 4, Aug. 2016, pp. 453463.
[3] X. Pan, B. Lee, and C. Zhang, A Comparison of Neural Network Backpropagation
Algorithms for Electricity Load Forecasting, IEEE Intelligent Energy Systems,Nov. 2013, pp.
2227
[4] Tiantian Ran, Songlin Sun, Bo Rong, Michel Kadoch, Game theory based multi-tier
spectrum sharing for LTE-A heterogeneous networks, IEEE International Conference on
Communications ,Sept. 2015
[4] Giovanni Giambene, Tara Ali Yahiya, LTE Planning for Soft Frequency Reuse, IFIP
Wireless Days (WD) ,Nov. 2013