4G Network Dimensioning

ITU ASP COE TRAINING ON

“Broadband Mobile and Internet of Things Network Planning”

Sami TABBANE
19-22 September 2017
Bangkok, Thailand

1
CONTENTS

Introduction
I. Dimensioning Methods
II. Traffic Evaluation
III. S1 U, S1 C and X2 Dimensioning
IV. Case Studies

2
Introduction

3
CIRCUIT SWITCHED VERSUS PACKET SWITCHED DIMENSIONING

Circuits Packets
Maximum bitrate / connection Fixed (limited) Variable (≤ maximum)
QoS (GoS, throughput, data loss …) Deterministic Unpredictable
Traffic models Simples Complexes
Simultaneous connections number Limited Adaptable and flexible
No optimization, possible Optimization with complex
Network resources
waste resource allocation algorithms
Resource management Simple Complex

4
TRAFFIC AGGREGATION

Packets

Dimensioning objective: determine the total required bandwidth

to carry the aggregated traffic with related QoS targets 5
DIFFERENT BURSTINESS LEVELS EXAMPLE AT SAME PACKET RATE

6
Traffic models

7
DATA TRAFFIC MODELING

• Data traffic is often modeled by:

– Packet size
– Packet inter-arrival time
• Some models:
– White random process,
– ARMA auto regressive moving average (Gauss),
– Markov-modulated processes,
– Fractional Brownian Motion (Long-range dependence),
– Wavelets (self-similar),
– TES (Transform Expand Sample) models (used for example to
model MPEG4 video traffic).

8
TRAFFIC MODELS EXAMPLES

QoS
Characteristics Model
Requirements
* Alternating talk- Delay < ~150 ms * Two-state (on-off) Markov
spurts and silence Jitter < ~30 ms Modulated Rate Process (MMRP)
intervals. Packet loss < ~1% * Exponentially distributed time at
Voice
* Talk-spurts produce each state
constant packet-rate
traffic
* Highly bursty traffic Delay < ~ 400 ms K-state (on-off) Markov
(when encoded) Jitter < ~ 30 ms Modulated Rate Process (MMRP)
Video
* Long range Packet loss < ~1%
dependencies
Interactive * Poisson type Zero or near-sero Poisson, Poisson with batch
* Sometimes batch- packet loss arrivals, Two-state MMRP
FTP
arrivals, or bursty, Delay may be
telnet important
web or sometimes on-off

9
 DATA SERVICES DIMENSIONING PROCESS

Classification of the traffic for priority handling

Applications Voice http ftp …

Video

Traffic classes
Class 1 traffic
…
Class 2 traffic Class 3 traffic

Traffic classes
considered for the Streaming traffic Elastic traffic
dimensioning

10
TRAFFIC VOLUME ESTIMATION

• Models integrated into a simulator which

generates, for different scenarios (e.g.,
service usage, subscriber types, codecs,
…) an aggregated traffic volume used to
dimension the capacity of the nodes
and/or the interfaces.
• Drawback: complex, time consuming and
requires accurate hypothesis.

11
QoS ASPECTS: EXAMPLE OF LTE QCI VALUES

12
Bandwidth based dimensioning

13
GENERAL APPROACH DESCRIPTION
• Dimensioning (UL or DL): based on the services
required bandwidth estimation.
• Contention ratios: to reflect the bursty nature of the
traffic and of the service activity as well as the
priorities of the users and services.
• Aggregation of the traffic flows bitrates: to estimate
the total link or node capacity.
• In case of overload due to unpredicted users and
services behavior: scheduling and queuing
mechanisms allow to maintain the QoS of high priority
traffic. Some services will see degradation of their QoS
parameters (e.g., bitrate, jitter, delay, BLER, …).
14
BURTSINESS, ACTIVITY RATE, CONTENTION RATIO
Busy Hour

Call Duration

Voice

Bitrate = 12,2 kb/s (ex.)

Burstiness = 0
Activity rate (e.g., 40%)

Busy Hour

Session Duration

Web
Average
Peak bitrate Contention ratio
Burstiness Provisioned bandwidth
Average bitrate
Activity rate (100%)

15
DEVICE QUEUING MECHANISMS

• Common queue examples for IP routers

– FIFO: First In First Out
– PQ: Priority Queuing
– WFQ: Weighted Fair Queuing
– Combinations of the above

• Service types from a queuing theory standpoint

– Single server (one queue - one transmission line)
– Multiple server (one queue - several transmission lines)
– Priority server (several queues with hard priorities - one transmission
line)
– Shared server (several queues with soft priorities - one transmission line)

16
 PRIORITY QUEUING
• Packets are classified into separate queues
– E.g., based on source/destination IP address,
source/destination TCP port, etc.
• All packets in a higher priority queue are served before a lower
priority queue is served
– In routers, if a higher priority packet arrives while a lower
priority packet is being transmitted, it waits until the lower
priority packet completes

17
 SERVICES CLASSIFICATION AND QUEUING EXAMPLE
Voice, visiophony RT
e.g., 20 ms, 12,2 kb/s
User 1
(voice) t

e.g., 384 kb/s Round Robin / WRR

User 2
(visio)
..
t
... ...
User i .
(voice) t

Email, FTP, … NRT

User 1
...
(email) t

User 2
PFS  Max(C/I)
(email)
..
t
...
.
User i
(FTP) t

18
 SERVICES CLASSIFICATION AND QUEUING EXAMPLE
Video streaming NRT

User 1 t

Round Robin / WRR

User 2 t
... ...
..
.
User i t

Web browsing NRT

...
User 1 t

PFS
User 2
..
t
...
.
User i t

19
SERVICES TRAFFIC AGGREGATION AND PRIORITIZATION EXAMPLE
Scheduler Priorities

Signaling 0

... ...
Non elastic traffic (jitter and delay)
1 Round Robin / WRR
... ... ...
Low elastic traffic (delay)
2 Round Robin / WRR Bearer
... ... ...

Elastic traffic
... n PFS
...

Very elastic traffic (BER)

m PFS  Max(C/I)
...

20
RADIO INTERFACE DIMENSIONING
Aggregated bearers

...

D?

...

Dimensioning purpose: determine the bearer bitrate D

21
THEORETICAL CELL AVAILABLE BANDWIDTH
Inter site distance
Cell load
500 m. 1732 m. Capacity = Ctotal*Loading Factor*Scaling Factor
0% 2.17384 1.68426
5% 2.16822 1.44588 User Maximum L1 Maximum
3GPP Maximum L1
10% 2.07655 1.25680 Equipment datarate number of
Release datarate Uplink
Category Downlink DL MIMO layers
15% 1.96951 1.16004
20% 1.86613 1.12126 Release 8 Category 1 10.3 Mbit/s 1 5.2 Mbit/s

25% 1.77003 1.08615 Release 8 Category 2 51.0 Mbit/s 2 25.5 Mbit/s

30% 1.67839 1.03514
35% 1.59381 1.03514 Release 8 Category 3 102.0 Mbit/s 2 51.0 Mbit/s

40% 1.51367 1.02943 Release 8 Category 4 150.8 Mbit/s 2 51.0 Mbit/s

45% 1.44084 1.01242
50% 1.37164 1.00934
Release 8 Category 5 299.6 Mbit/s 4 75.4 Mbit/s

55% 1.30975 1.00622 Release 10 Category 6 301.5 Mbit/s 2 or 4 51.0 Mbit/s

60% 1.25242 1.00491
Release 10 Category 7 301.5 Mbit/s 2 or 4 102.0 Mbit/s
65% 1.20192 1.00328
70% 1.15599 1.00211 Release 10 Category 8 2998.6 Mbit/s 8 1497.8 Mbit/s
75% 1.11735 1.00000
CTotal
80% 1.08236 1.00000
85% 1.05467 1.00000
90% 1.03153 1.00000 Example: Ctotal = 150.8 Mb/s, Loading Factor = 60%
95% 1.01523 1.00000 and Scaling Factor = 1.25242 (for cell radius = 500m.)
100% 1.00000 1.00000 Capacity = 150.8*0.60*1.25242 = 113.319 Mb/s
Loading factor 22
 DIMENSIONING PROCESS
Step 1: initial configuration (dimensioning and planning)

Traffic and mobility model

Subscribers classes
Services QoS characteristics Aggregate required capacity (Mb/s)
Services and subscribers contention ratios
Transmission and protocols overheads Initial number of
Theoretical cell available bandwidth required cells

Radio planning
Radio interface characteristics
Coverage and interference requirements Cell available bandwidth (Mb/s)
Subscribers geographic distribution
Allocated spectrum

23
DIMENSIONING PROCESS
Step 2: final configuration

Cell available bandwidth (Mb/s)

Aggregate required capacity (Mb/s)
Final number of required
Coverage and interference characteristics
cells and sites configuration
Radio interface characteristics
New radio planning: optimization

24
INPUTS

• Subscribers classes,
• Service usage/subs. class,
• Contention ratios/subs. class,
• Subscribers geographic distribution,
• Services bitrates,
• Services and protocols overheads.

25
EXAMPLE AUDIO ENCODING TECHNIQUES

G.711 G.729
• PCM (non-linear) • CS-ACELP
• 4KHz bandwidth • 4KHz bandwidth
• 64Kb/s • 8Kb/s
G.723.1
G.722
• MP-MLQ
• SB-ADPCM
• 4KHz bandwidth
• 4-8KHz bandwidth • 5.3/6.3Kb/s
• 48/56/64Kb/s GSM
G.728 • RPE/LTP
• LD-CELP • 4KHz
• 4KHz bandwidth • 13Kb/s
• 16Kb/s

26
EXAMPLE VIDEO ENCODING TECHNIQUES

MPEG1
• Up to 1.5Mb/s
MPEG2 H.261 and H.263
• Up to 10Mb/s (HDTV • n  64Kb/s, 1 n  30
quality)
MPEG4
• 5-64Kb/s (mobile,
PSTN)
• 2Mb/s (TV quality)
• MPEG7, MPEG21

27
MCS DESCRIPTION AND REQUIRED RECEIVER SENSIBILITY

28
MSC DISTRIBUTION IN THE CELL EXAMPLE

64-QAM

29
DISTRIBUTION OF THE CAPACITY PER QOS TYPE

30
 CONTENTION RATIO
• Measures the simultaneity of users requesting bit rate from the BS
because most users won’t demand data at the same time. The
absolute peak demand on shared resources rarely occurs. User
simultaneity is defined by the contention ratio.
• If many of the connected subscribers demand data, their packets
will be delivered assuming some latency or jitter (less priority).
• Example: if 2 contention ratios are defined for the non-guaranteed
partition of the bandwidth (e.g., 30 for residential users (less
priority) up to 10 for business users (higher priority and
throughput)), we have:
Subscriber class DL BE service Offered data rates
Residential Class X = 512 kbps X’ = 512/30=17 kbps
Business Class Y = 1 Mbps Y’ = 1000/10=100 kbps
 Actual data-rates considered in the system capacity calculations.

31
 OVER SUBSCRIPTION RATIO (1)

• OSR = ratio of the total subscriber’s demand over the

reference capacity of the base station when taking into
account the adaptive modulation.
• The reference capacity of the base station corresponds to
the available bit rate of the lowest modulation scheme
served with that BS (here BPSK1/2).
Cref = FFTused/2Ts
• FFTused and Ts (symbol duration) values depend on the
channel bandwidth (in LTE: from 72 to 1200) and the
Cyclic Prefix factor respectively (=CP/Symbol duration).

32
OVER SUBSCRIPTION RATIO (2)

• Residential class = A% of the users in the cell,

• Business class = B% of the users in the cell.
• Total capacity for OSR calculation:
Ctot = N*(A%*X + B%*Y)
OSR = Ctot/Cref

Worldwide distribution of
service classes

33
DL CHANNEL RAW BANDWIDTH

• FFTused = number of data subcarriers dependent

on the channel bandwidth (from 72 to 1200 in
LTE).
• %P = percentage (weight),
• k = number of bits per symbol,
• OCR = overall coding rate.

34
APPLICATION BANDWIDTH AND AVERAGE USAGE

Weight (average
Data rate
Application usage during a
(kbps)
session)
Multiplayer interactive gaming D1 = 50 W1 = 25.0 %
VoIP and Video Conference D2 = 32 W2 = 10.0 %
Streaming Media D3 = 64 W3 = 12.5 %
Web browsing and instant
Nominal W4 = 32.5 %
messaging
Media content downloading BE W5 = 20.0 %

35
CONTENTS

I. Dimensioning Method

36
DIMENSIONING METHOD

Planning Design: Application to LTE

37
PLANNING DESIGN

 LTE dimensioning requires three steps:

 C/I Distribution
 Subscribers
Distribution
 Subs Number / Class Traffic  Radio Planning and
 Traffic Model Evaluation Link budget

Bandwidth  Cell Average

Capacity

S1 U - S1 C -X2

A GW – P GW

38
CONTENTS

II. Traffic Evaluation

1. Subscribers population
2. Mobility and Traffic Model
3. Cell Characteristics and Planning assumptions

39
TRAFFIC ESTIMATION

1. Subscribers population

40
SUBSCRIBERS POPULATION

 Inputs - Outputs

Input Subscribers (A) Subs. Number/Class

(1) Population (in Thousands)

(2) Market Share Estimation (%) (B) Subs. Density
(3) Subscribers Classes (VIP, SME,… )
(4) BH Session Number
(5) Area (in Km2)
(6) Subscriber Growth Rate (%)
(7) Planning Period (in years) (C) Growing Evolution

41
SUBSCRIBERS POPULATION

Formulas

• (A) Subs. Number/Class = (1) x (2) x (3)

• (B) Density / Subs. Class = (A) ÷ (5)

• (C) Subs. Growth for n year(s) = (B)[ 1+ (6)] (7)

42
TRAFFIC ESTIMATION

2. Traffic Model

43
TRAFFIC MODEL
 Inputs – Outputs
Mobility & Traffic Model
(D) Throughput/Service
(8) Service Average Bitrate (kb/s)
(9) Service Activation Number at BH per
(E) Throughput/Subs
Subs Class
(10) Service Activity Rate
(11) Burstiness Margin / Service (F) Throughput/Subs Class
(12) Service or Signaling Procedure Average
Duration
(13) Average BLER (for NRT services) (G) Bandwidth/Area
(14) Contention Ratio / Service / Subs. Class
(15) TAUs number / Subscriber at BH
(H) Bandwidth / Evolution
(16) HO number / Subscriber at BH
(17) Mobility Margin (for the area)
(18) Security Margin (for total traffic (I) Traffic /Area
volume). Here taken instead of OSR
(19) Signaling bitrate / Procedure
(J) Traffic /Area Growing
(20) Low layers overhead
(21) Carrier Bandwidth (MHz)
44
TRAFFIC MODEL
Formulas

• (D) Service bandwidth = (8) x (10) x [ 1 + (11)] x [ 1 + (13)] x [ 1

+ (20)]
• (E) Throughput / Subscriber class = (D) x [(9) x (12) x (14) /
3600] + (19) x [((4), (15), (16)) x (12) / 3 600]
• (F) Bandwidth/ area = [∑ (A) x (E)] x [ 1 + (17)] x [ 1 + (18)]
• (G) Bandwidth Evolution = ∑ (C) x (F)
• (H) Bandwidth / km2 = (F) / (5)

45
TRAFFIC MODEL

Assumptions

• (K) Initial Cell Capacity value (Starting Assumption) = Peak

Rate (100 Mb/s for LTE) x (21) / 20 MHz (for LTE) x 70%

• (M) Number of Cells (First estimation) = (F)/(K)

46
TRAFFIC ESTIMATION

3. Cell Characteristics and Planning assumptions

47
CELL CHARACTERISTICS AND PLANNING ASSUMPTIONS
 Link Budget principle

48
CELL CHARACTERISTICS AND PLANNING ASSUMPTIONS

 Planning Process

Initial Planning

1. Model Tuning
Planning
2. User density and distribution
3. Site Selection
4. eNB number (M) Coverage Planning
5. Configuration Planning (Tilt, Optimization
Azimuth, Height,…)
1. Coverage Rate
2. Parameters Tuning
3. Interference Rate

49
 CELL CHARACTERISTICS AND PLANNING ASSUMPTIONS

 Inputs - Outputs

(M) Number of eNB before (Q) subs/ MCS type area

planning (dimensioning output) distribution
Cell Characteristics and
Planning Assumptions
Step 2
(P) Throughput UL - DL
(22) MCS Types
 Radio Planning (23) MIMO
 Link Budget Configuration
(24) Distribution of the
(O) Cell Capacity
subs in the area (%)
(25) Coverage Rate

Step 1
(M’) Number of cells

50
CELL CHARACTERISTICS AND PLANNING ASSUMPTIONS
 Coverage estimation

(22) Bandwidth (23) MIMO (24) Distribution of

(21) MCS f[(21), (22), (23)}
(MHz) Configuration the subscribers
QPSK 1/8 x x
QPSK 1/5 x x
QPSK 1/4 1,4 SISO x x
QPSK 1/3 x x
3 SIMO
QPSK 1/2 x x
5 MISO
QPSK 2/3 x x
QPSK 4/5 10 MIMO x x
16 QAM 1/2 x x
15
16 QAM 2/3 x x
20
16 QAM 4/5 x x
64 QAM 2/3 x x
64 QAM3/4 x x
64 QAM 4/5 x x

(O) Cell Capacity = ∑ (24) x f{(21), (22), (23)}

51
PEAK DATA RATES DL AND UL (IN LTE)

52
 CELL CHARACTERISTICS AND PLANNING ASSUMPTIONS
 Numbers of sites

Sites Evaluation (R) Number of cells

(F) Traffic / Area

(O) Cell Capacity
(25) Sectorisation Type
(P) Number of Sites

(R) Number of cells = (F) / (O)

(P) Number of sites = (R) / (25)

If (R) ≠ (M’) new radio planning process is required. (O) may change and the new (R) may
also change.
Iterative process is then required until (R) = (M’).

53
CONTENTS

III. S1 U, S1 C and X2 Dimensioning

1. Bandwidth Inputs-Outputs
2. Traffic Model
3. Formulas

54
S1 U, S1 C AND X2 DIMENSIONING

1. Bandwidth Inputs-Outputs

55
S1 U, S1 C AND X2 DIMENSIONING
 Inputs - Outputs
S1 U Link
(22) Service (i) Use (T) S1 U BW
(23) Service Overhead
(24) % of subs

S1 C Link
(25) Number of TAUs (U) S1 C BW
(26) Number of HO
(27) Mobility Margin

3% (S) (V) X2

Link Criteria
(28) Link bandwidth (W) Link Number
(29) Support Bandwidth
(30) Usage Rate

56
S1 U, S1 C AND X2 DIMENSIONING

2. Traffic model

57
S1 U, S1 C AND X2 DIMENSIONING
 Codecs Model

Radio Interface Radio Interface

Compression
Traffic Model Codecs with HO without HO
Ratio
compression compression
VoIP AMR x x x
Video
MPEG4 x x x
Conference
HTTP x x x x
Web Browning x x x x
FTP x x x x

Video Streaming MPEG4 x x x

email x x x x
Interactive
x x x x
Gaming

58
S1 U, S1 C AND X2 DIMENSIONING

 Traffic Model on S1 U link

Bit Rates (23) IP Overhead

(22) Traffic Model Codecs (24) % of use
(Kbps) (byte)

VoIP AMR 12,2 40 50 %

Video Conference MPEG4 64 1 000 30 %
HTTP x 0,576 1071 60 %
Web Browning x 1,209 600 20 %
FTP x 2 000 2 000 10 %
Video Streaming MPEG4 64 400 10 %
email x 80 22,7 10 %
Interactive Gaming x 128 300 5%
P2P 500 500 20%

59
S1 U, S1 C AND X2 DIMENSIONING

 Link Mobility Model on S1 C link

(25) Number of TAUs 500

(26) Number of HO 475

60
S1 U, S1 C AND X2 DIMENSIONING

3. Formulas

61
S1 U, S1 C AND X2 DIMENSIONING
Formulas

• (T) S1 U BW = ∑ 8kb/s x (23) x (24)

• (U) S1 C BW = 8 kb/s [(25) + (26))]

• (V) "X" _"2" = 3% S1 BW

• (W) Number of Links = (28) / [(29)/(30)]

62
CONTENTS

IV. Case Studies

63
TRAFFIC MODEL EXAMPLE

UL DL
Traffic PPP Through- PPP Through-
BHSA Bearer PPP Bearer PPP
Parameters Session put per Session put per
rate Session BLER rate Session BLER
Duty User Duty User
(kb/s) Time (s) (kb/s) Time (s)
Ratio (kbps) Ratio (kbps)

VoIP 1.4 26.9 80 0.4 1% 0.34 26.9 80 0.4 1% 0.34

Video
0.2 62.52 70 1 1% 0.25 62.52 70 1 1% 0.25
phone
Video
0.2 62.52 1800 1 1% 6.32 62.52 1800 1 1% 6.32
Conference

RT Gaming 0.2 31.26 1800 0.2 1% 0.63 125.05 1800 0.4 1% 5.05

Streming
0.2 31.26 1200 0.05 1% 0.11 250.11 1200 0.95 1% 16.00
Media
IMS
5 15.63 7 0.2 1% 0.03 15.63 7 0.2 1% 0.03
Signaling
Web
0.6 62.52 1800 0.05 1% 0.95 250.11 1800 0.05 1% 3.79
Browsing

Ftp 0.3 140.68 600 1 1% 7.11 750.33 600 1 1% 37.90

Email 0.4 140.68 50 0.5 1% 0.39 750.33 15 0.3 1% 0.38

P2P file
0.2 250.11 1200 1 1% 6.73 750.33 1200 1 1% 6.73
sharing 64
TRAFFIC MODEL EXAMPLE (2)

UL DL UL DL
Traffic PPP PPP Through- Through-
Parameters Bearer PPP Bearer PPP
Session Session put per put per
rate Session BLER rate Session BLER
Duty Duty Session Session
(kb/s) Time (s) (kb/s) Time (s)
Ratio Ratio (kbit) (kbit)

VoIP 26.9 80 0.4 1% 26.9 80 0.4 1% 869 869

Video
62.52 70 1 1% 62.52 70 1 1% 4 421 4 421
phone
Video
62.52 1800 1 1% 62.52 1800 1 1% 113 687 113 687
Conference

RT Gaming 31.26 1800 0.2 1% 125.05 1800 0.4 1% 11 369 90 950

Streming
31.26 1200 0.05 1% 250.11 1200 0.95 1% 5 684 864 023
Media
IMS
15.63 7 0.2 1% 15.63 7 0.2 1% 22 22
Signaling
Web
62.52 1800 0.05 1% 250.11 1800 0.05 1% 5 684 22 737
Browsing

Ftp 140.68 600 1 1% 750.33 600 1 1% 85 265 454 749

Email 140.68 50 0.5 1% 750.33 15 0.3 1% 7 105 11 369

P2P file
250.11 1200 1 1% 750.33
65 1200 1 1% 303 166 909 498
sharing
TRAFFIC MODEL EXAMPLE (3)

Dense Urban
Traffic BH Throughput Per User
User Behavior
Penetration BHSA (bps)
Ratio UL DL
VoIP 100% 1.4 335 335

Video Phone 20% 0.2 243 243

Video Conference 20% 0.2 6 252 6 252

RT Gaming 30% 0.2 625 5 002

Streaming Media 15% 0.2 104 15 840

IMS Signaling 40% 5 30 30

Web Browsing 100% 0.6 938 3 752

FTP 20% 0.3 7 034 37 517

Email 10% 0.4 391 375

P2P file sharing 20% 0.2 16 674 50 022

Total 32 626 119 368

Balance Dimensioning Parameters

Peak to Average Ratio 40%
UL BH Throughput per User (bps) 10 802
DL BH Throughput
66 per User (bps) 44 771
Number of Launched Subscribers 1 000 000
CELL AVERAGE THROUGHPUT BASELINE

Cell Average Throughput DL/UL (Mpbs) @20

Scenario MHz BW
2.6 GHz
Dense Urban 34.3/19.8
Urban 34.2/19.8
Suburban 26.3/14.0
Rural 26.3/14.0
Assumptions:
 Standard hexagon cellular structure,
 19 sites, 3 cells/site,
 ISD 500m in Dense Urban and Urban scenarios,
 ISD 1700m in Surburban and Rural scenarios,
 Frequency reuse: 1x3x1,
 DL 2X2 CL switch (rank1/rank2),
 UL 1X2 IRC.
67
Thank you!

68