Crisan Sergiu Vasile Tst Eng 2019 [611713]

1
Content
1. Contents ……………………………………………………………… …… ……1
List of a bbreviations ………………………………………………… …………. .3
2. Rezumat în limba română ………. …………. …………………………………… ………….. .6
2.1 Stadiu actual …………………………………………………………………… ………………. .6
2.2 Fundamentare teoretic ă ……………………………. ……………………… ………………. .6
2.3 Implementare ………………………. …………………………………………. ………………. 9
2.4 Rezultate experimentale ………………………….. ……………………….. ………………. 9
2.5 Concluzii ………………………………………………………………………… …………….. 13

Work Planning ……………………………………………… …… ………… .….14

3. State of the art ….……………………………………………………… ……… .15

4. Theoretical fundamentals ………………… ..…….………….……… …………18
4.1 Introduction ……… ……………………………………………. ………….. .18
4.2 LTE network architecture overview …………………………… …………….18
4.3 Detailed breakdown of LTE architecture ……………………….. ……………19
4.3.1 Evolved packet core…….. ……………………………. .………… ….20
4.3.2 Evolved universal terrestrial radio access network ……..…………… .21
4.3.3 The user equipment ……. …………… .……. ………… …………… …21
4.4 LTE physical layer …… …… …………………………………….. ……….. .22
4.4.1 OFDMA and SC -FDMA ………………………………… …………. .22
4.4.2 Pathloss models ………………………………………… ………… ….23
4.4.3 Frequency reuse algorithm …..………………………… ………… …..24
4.4.4 MIMO Technique ………………………………………… ………… .24
4.5 LTE radio protocol architecture …………………………………… ……… .25
4.5.1 User plane ………………………………………………… …………….. .25
4.5.2 Control plane ……………………………………………. .…………. .26
4.6 Quality of Service ……… ………………………………………… ……… .27
4.6.1 MAC schedulers ………………………………………… …………. ..28
4.7 Carrier Aggregation ……………………………………………… ……… .29
4.8 Handover procedure ……………….. ………………………… …… ……… 29
4.9 Link adaptation ….…………………………………………………………31

2

5. Implementation ………………………………………………………. .…….. 32
5.1 Introduction to NS-3 ……………………………………………… ………. 32
5.2 Handover simulation using NS-3 …………………………………. ……… 36
5.2.1 Study of the X2 handover procedure …………………… ……….….. 36
5.3 Carrier Aggregation simulation using NS -3 ………………………. ……… 38
5.3.1 Study of carrier aggregation on stationary nodes ……………… .……..39
5.3.2 Study of carrier aggregation on moving nodes ……………… ……… .40
5.4 MAC schedulers simulation in NS -3 ….…………………………… ………. .43
5.4.1 Study of scheduling mechanism on non -QoS aware schedulers ………43
5.4.2 Study of scheduling mechanism on QoS aware schedulers …… ..….…45

6. Experimental results ………………… .…………………… …………… ……47
6.1 Handover simulation results ………………………… ..…………………… 47
6.2 Carrier Aggregation simulation results ………… …… .………………….. 50
6.2.1 Carrier Aggregation results on stationary UE ………… ……… ..…….50
6.2.2 Carrier Aggregation results on moving UE ……… ..…….. ……………… 53
6.3 MAC Scheduling mechanism results …… ..………. ………………… .…… 54
6.3.1 MAC scheduling results on non -QoS aware schedulers ……………… 54
6.3.2 MAC scheduling results on QoS aware schedulers ……. ………………… 56

7. Conclusion s ………………………………………………… ……….. ……… .59

8. References …………… …………………………………………… ……….. ….60

9. Appendix ……………………… ….……………………………… …………… 61
9.1 NS -3 class diagram……………… ..………………………………………. 61
9.2 NS-3 handover simulation source code ………………………………… …62
9.3 NS-3 carrier aggregation simulation source code ……… ……………… …70
9.4 NS-3 MAC scheduling mechanism simulation source code …… ……… …76
9.5 Curriculum Vitae …………………………………………………… ..…… 84

3
List of a bbreviations
GSM Global System for Mobile Communications
UMTS Universal Mobile Telecommunications System
LTE Long Term Evolution
UTRAN UMTS Terrestrial Radio Access Network
E-UTRAN Evolved UMTS Terrestrial Radio Access Network
EPS Evolved Packet System
SAE System Architecture Evolution
EPC Evolved Packet Core
RAN Radio Access Network
FDD Frequency Division Duplex
TDD Time Division Duplex
CS Circuit Switching
PS Packet Switching
IP Internet Protocol
VoIP Voice Over IP
eNB Evolved Node B
PDN Packet Data Network
QoS Quality of Service
UE User Equipment
MME Mobility Management Entity
S-GW Serving Gateway
P-GW PDN Gateway
HSS Home Subscriber Server
PCRF Policy Control and Charging Rules Function
TTF Traffic Flow Template
GBR Guaranteed Bit Rate
CN Core Network
NAS Non-Access Stratum
OFDM Orthogonal Frequency Division Multiplexing
OFDMA Orthogonal Frequency -Division Multiple Access
SC-FDMA Single Carrier Frequency Domain Multiple Access
RRM Radio Resource Management
ME Mobile Equipment
MT Mobile Termination
TE Mobile Termination
UICC Universal Integrated Circuit Card
USIM Universal Subscriber Identity Module
PSK Phase -shift keying
QAM Quadrature amplitude modulation
QPSK Quadrature Phase Shift Keying
RF Radio Frequency
LOS Line of Sight

4
ICI Inter -Cell Interference
SFFR Strict Fractional Frequency Reuse
SFR Soft Fractional Reuse
MIMO multiple -input and multiple -output
SNR Signal to Noise Ratio
TCP Transmission Control Protocol
UDP User Datagram Protocol
RRC Radio Resource Control
PDCP Packet Data Convergence Protocol
RLC Radio Link Control
MAC Medium Access Control
GTP GPRS Tunneling Protocol
SDU Service Data Unit
PDU Protocol Data Unit
HARQ Hybrid automatic repeat request
TM Transparent Mode
UM Unacknowledged Mode
AM Acknowledged Mode
SN Sequence Numbers
QCI QoS class Identifier
ARP Allocation and Retention Priority
DCI Data Control Indication
MCS Modulation Coding Scheme
CQI Channel Quality Indicator
RB Resource Blocks
RBG Resource Block Group
FD Frequency Domain
TD Time Domain
TBR Target Bit Rate
CQA Channel and QoS Aware
HOL Head of Line
CC component carriers
CA Carrier Aggregation
PHY Physical Layer
S1-AP S1- Application Protocol
X2-AP X2- Application Protocol
PCC Primary Component Carrier
SCC Secondary Component Carrier
RSRP Reference Symbols Received Power
RSRQ Reference Symbols Received Quality
RSSI Received signal strength indication
SINR Signal to Interface Noise Ratio
AMC Adaptive Modula tion and Coding

transmission power [W]

5
reception power [W]
transmission gain (unit -less)
reception gain (unit -less)
Wavelength [m]
distance [m]
system loss (unit -less)
𝐻𝑡 Height transmitter [m]
𝐻𝑟 Height receiver [m]
three distance fields [m]
path loss at reference distance [dB]
distance [m]
path loss [dB]
frequency [MHz]
eNB height above the ground [m]
UE height above the ground [m]
is a logarithm in base 10 (this for the whole document)

6
2. Rezumat în limba română
2.1 Stadiul actual
Scopul principal al unui simulator este de a construi un model virtual a unui sistem complex
precum o rețea LT E sau alte sisteme complexe . Motivul care stă la baza utilizări i simulatoarelor în
locul testări pe un sistem real este că sistemele reale sunt scumpe, complexe, indispo nibile pentru
scopuri academice sau cercetare. Simulatorul ns-3 este un simulator de rețea bazat pe evenimente
discrete , destinat în principal pentru cercetare și utilizare in scopuri educațional e. NS -3 este o
platformă “open source ”, creată de comunitatea de cercetare .
Scopul simulatorului ns -3 acestui software este de a furniza mijloacele necesare pentru
efectuarea cercetări și de a fi o platformă extensibilă de simulare a sistemelor complexe . Unele
dintre caracteristicile oferite sunt modelele pentru funcționarea și măsurarea performanț ei rețele lor
de date cu comutație de pachete , posibilitatea de a simula diverse experimente în studiul
comportamentului sistem elor în anumite scenarii .
2.2 Fundamentare teoretică
LTE a evoluat de la predecesorul său UMTS, o consecință a acestui fapt este că a păstrat o
anumită asemănare cu predecesorul său . Arhitectura LTE este o arhitectură în care utilizatoru l
comunic ă direct între nucleul rețelei și rețelele radio de acces. Această schimbare de arhitectură
poate fi văzută în imaginea de mai jos.

Figur a 2.1 – Comparație între arhitectur a UMTS si LTE

Arhitectura LTE poate fi rezumată la un singur termen si anume EPS . EPS este un termen care
se referă la întregul sistem de la un capăt la altul , care este compus din UE, E -UTRAN și EPC. EPS
este un sistem care funcționează cu conectivitate IP ca și o rețea de date cu comutație de pachete
pentru a oferi utilizatorului acces la internet. Această structură este utilizată și pentru majoritatea
serviciilor oferite de LTE . Structura sistemului EPS este prezentata in imaginea de pe pagina
următoare .

7

Figur a 2.2 – Schem a bloc a sistemului EPS

Arhitectura LTE este compusă din următoarele trei componente principale:
• UE, echipamentul utilizatorului.
• E-UTRAN, rețeaua de acces radio
• EPC, denumit și nucleul rețelei

EPC conține o mare parte a infrastructurii care se ocupă cu autentificarea, gestionarea
mobilității împreună cu eNB prin procesul de handover , rutarea traficului, autodiagnosticarea și
întreținerea sistemului . Nucleul rețelei este în principal compus din MME, S -GW și P -GW. EPC
cuprinde și alte noduri logice precum : HSS, PCRF.
E-UTRAN poate fi considerată o componenta esențială a structurii EPS, spunem acest lucru
deoarece efectuează o singură sarcină, aceea de a conecta UE cu restul rețelei centrale. E -UTRAN
ca restul rețelei LTE împărtășește unele dintre caracteristicile sale cu UMTS, această similitudine
poate fi observată în schema de denumire. Pe scurt, E -UTRAN este alcătuit dintr -o rețea de eNB –
uri formată din interfețe standardizate după cum poate fi văzut in imaginea de mai jos.

Figur a 2.3 – Exemplu a unei rețele E -UTRAN

Echipamentul mobil (UE) reprezintă o parte a arhitecturii LTE cu care majoritatea oamenilor
interacționează zilnic, din punct de vedere al unei persoane obișnuite, acesta este singur a
componentă a rețelei pe care o cunoaște .
Comunicarea dinte UE si rețeaua de eNB -uri se face prin intermediul OFDM în LTE, deși
OFDM nu este utilizat în forma sa originală, putem vedea totuși că variantele sale sunt utilizate,
adică OFDMA pentru transmisia de la rețea la utilizator și SC -FDMA pentru transmisia de la
utilizator la rețea .
La fel ca în orice altă rețea de telefonie mobilă , deficitul de lățime de bandă, spectrul limitat
de frecvențe radio și limitările de putere au creat necesitatea folosiri unui mod eficient de utilizare
a lățimii de bandă . În LTE există două tipuri de algoritmi de refolosire a frecvenței: SFFR si SFR .
O altă tehnologie utilizata cu succes in LTE este MIMO . Această tehnologie utilizează mai
mult de o antenă la un moment dat pentru a oferi performanțe sporite. În prezent, în LTE există mai
multe moduri de transmisie care pot fi utilizate în timpul utilizării MIMO, unele dintre ele fiind:

8
• Antena unică, o singură antenă care transmite un singur flux de date și este recepționată de
una sau mai multe antene
• Diversitate la transmisie : transmite un flux de informații folosind mai multe antene
• Multiplexarea spațială cu buclă deschisă: această formă de MIMO trimite două fluxuri de
informații care sunt transmise pe două sau mai multe antene. Nu există feedback de la UE.
• Multiplexarea spațială în buclă închisă: este ca o versiune cu buclă deschisă, dar are un
mecanism de feedback integrat.

Un important aspect al rețelei LTE este a rhitectura protoco alelor radio , aceasta poate fi
împărțită în două straturi separate, acesta fiind planul de control și planul utilizatorului .
În planul utilizatorului, aplicația creează pachete de date care sunt trimise prin protocoale
precum TC P, UDP și IP. Pe cealaltă parte, planul de coman dă sau control are protocolul RRC,
care scrie mesajele de semnalizare referitoare la aspecte precum calitatea legături radio, starea
celulelor învecinate și tehnologia de acces radio care sunt schimbate între eNB și UE. Pentru
ambele planuri, informațiile sunt prelucrate de protocoalele PDCP, RLC și MA C, înainte de a fi
transmise stratului fizic.
O funcție important ă in LTE este QoS, care se realizează cu ajutorul purtătoarelor r adio, care
sunt furnizate de către MME la eNB utilizând atributele QoS standardizate. Pe baza acestor atribute
straturile EPS și protoco alele între UE și eNB pot gestiona traficul spre rețea și dinspre rețea . În
general, servicii le de purtător pot fi clasifica te în două categorii, în funcție de natura nivelului QoS
setat pent ru purtătorul radio :
• Purtătorii de tip GBR aceștia au stabilit o un număr de resurse fix alocate pe tot parcursul
conexiuni
• Purtători de tip non-GBR, acești purtători nu garantează o valoare fixă a debitului . Acestea
sunt folosite pentru aplicații cum a r fi transferul FTP și navigarea pe Internet. Resursele
nefiind alocate permanent.

O component ă vitala in funcționarea corecta a rețelei LTE in funcție de QoS este stratul MAC ,
acesta este prezent in UE și în eNB și este un element vital în funcționarea corectă a acestor două
entități cu restul rețele i si in organizarea resurselor in funcție de parametrul QoS.
O altă caracteristică importanta si relativ noua a rețelei LTE este agreg area purtătoare lor,
aceasta permite utilizarea mai eficientă a spectrului fragmentat. Agregare a purtătoarelor este
proiectată să utilizeze blocuri de frecvențe multiple, numite “component carrier ”, care sunt atribuite
unui utilizator. Aceast ă nouă funcționalitate este concepută pentru a crește viteza de transfer a
datelor și de echilibrare a sarcinii în rețea .
Procedura de handover în LTE joacă un rol important în funcționarea sa ca rețea. Ca și în
rețelele predecesoare rețelei LTE, problema mobilității prezintă mari pr ovocări, mai mult decât
oricând, deoarece LTE este prima rețea concepută pentru a fi o rețea mobilă IP.
În LTE, exista numai handover direct . Principalele obiective ale procedurii de handover în
LTE sunt menținerea nivelului QoS, mențin erea nivelul minim al consumului de energie astfel
încât bateria dispozitivului să nu fie drenată, UE ar trebui să poată continua să utilizeze serviciile
înainte și după procedura de handover , să poată efectua transferuri î ntre alte rețele și tehnologii
precum 2G, 3G. Procedura de handover se bazează pe măsurători efectuate de UE la parametri
precum RSRP, RSRQ si SINR.
Modularea și codarea adaptivă (AMC) este folosită în procesul de adaptarea legăturii care este
utilizat pen tru îmbunătățirea capacității sistemului . AMC este o tehnică care adaptează schema de
modulare și codare (MCS) în conformitate cu calitate a canalului radio de comunicație raportat ă de
UE.

9
2.3 Implementare
Prim a component ă din rețea ua LTE implementată pentru a fi simulată a fost procedura de
handover intre eNB bazat pe interfața X2. Pentru testarea în condiții apropiate de cele găsite în
lumea reală, au fost selectate mai multe modele de propagare, acest e model e de propagare variază
de la s implu l a complex. Aceste modele de propagare sunt: modelul de propagare in spa țiul liber a
lui Friis, modelul de propagare two ray si three log distance .
Configura ția folosită pentru rețea este compusă din 2 eNB -uri, având o distanță de 1000 de
metri între ele conectat e intre ele cu o interfață X2, EPC -ul rețelei LTE și un UE care se deplasează
între cele două celule cu o viteză de 20 m / s.
Cel de al doilea element de rețea studiat este simularea agregării purtăto arelor utilizând
suportul oferit de ns -3. Simularea agregării purtăto arelor a fost făcută în două etape, la început
topologia rețelei este formată dintr -un eNB și un UE staționar . UE este plasat mai întâi relativ
aproape, la o distanță de 100 de metri de eNB, du pă care este mutat mai departe de eNB la o distanța
de 400 de metri. În studiul CA pe nodurile staționare au fost adăugate diferite moduri de transmisie
și prin aceasta ne referim la MIMO , distanta la care a fost plasat UE -ul este de 1500 metri de eNB .
Modurile de transmisie incluse în scenariu sunt: anten ă unică și diversitate la transmisie ,
multiplexarea spațială cu buclă deschisă și închisă.
A doua parte în simularea agregării purtăto arelor a fost plasarea echipamentul ui utilizatorului
în apropierea e NB și apoi mișcării acestuia cu o viteză constantă către marginea celulei, aceasta a
fost făcută pentru a ilustra modul în care debitul scade în raport cu creșterea distanței și vitez a
dispozitivului . Această topologie este asemănătoare cu prima topologie, singura diferență fiind că
UE se deplasează cu o viteză constantă. Testele au fost efectuate la trei vitez e de deplasare,
respectiv 10 m / s, 20 m / s, 30 m / s.
A treia componentă a rețelei LTE studia tă în această teză sunt mecanismele de ordonare
implementate la nivel MAC . Studiul mecanismelor de ordonare se face pe câ țiva algoritmi de
ordonare și anume: round robin , maximum throughput , proportional fair , priority set scheduler și
channel and QoS aware scheduler sau CQA .
Topologia rețelei pentru acest studiu a constat dintr -un singur eNB și 3 UE, aranjat e într-o
manieră care simul ează diferite moduri de funcț ionare . Testarea pentru toți cei 5 algoritmi a fost
efectuată cu aceleași caracteristici de rețea și în aceeași configurare a rețelei.
Modelele de propagare folosite sunt modelul de propagare Okumura și model de propagare
spațiu liber . Testarea planificatorului MAC a fost efectuată în două etape. În prima etapă, toate
dispozitivele au avut același nivel de QoS, iar algoritmi utiliza ți pentru testarea a fost round robin ,
proportional fair și maximum throughput . În cea de -a doua etapă, UE aveau un nivel diferit de
QoS, mai exact un UE avea un nivel QoS non-GBR, iar celelalte două UE aveau un nivel QoS
GBR , iar alogoritmi folosiți au fost: round robin , priority set scheduler și channel and QoS aware
scheduler .
2.4 Rezultate experimentale
Acest capitol trece în revistă rezultatele experimentale obținute prin studierea mecanismului
de handover pe interfața X2, a conceptului de agregare a purtătoarelor și a mecanismului de
planificare pentru MAC cu ajutorul ns -3.
Primele rezultate experimentale sunt pentru algoritmului de handover pentru evenimentele A2
și A4 rezultatele fiind împărțite pe afișarea SINR, RSRP și MCS pentru downlink și uplink cu un
model de propagare adăugat , three log distance .

10

Figur a 2.4 – Rezultate pentru evenimentele A2 și A4 în downlink

La început, când UE începe să se deplaseze spre primul eNB, valorile RSRP și SINR cresc pe
măsură ce UE se depărtează de eNB încep să scadă până când procedura de handover este finalizat ă,
după care acestea încep să crească din nou până când UE începe să se îndepărteze de al doilea eNB.
Pentru graficul MCS, comportamentul descris de grafic este similar cu cel descris anterior pentru
graficele RSRP și SINR.

Figur a 2.5 – – Rezultate pentru evenimentele A2 și A4 în uplink

Imaginea de mai sus este axată pe statisticile prezente obținute pentru uplink . Comportamentul
descris în ele este similar cu ceea ce a fost descris pentru înainte cu diferența că valorile sunt mai
mici pentru RSRP și SINR, deoarece puterea de transmisie a UE este mai mică decât cea a eNB.
Deoarece puterea de transmisie a UE este mai mică, UE trebuie să schimbe schema de co dare și
modulație mai des decât eNB. Același comportament a fost observat si pentru algoritmului de
transfer al conexiuni bazat pe evenimentul A 3.
Următoarele rezultate sunt pentru agregarea purtătoarelor și acoperă modul în care viteza de
transfer a unui UE variază atunci când agregarea purtătorului este activată comparativ cu agregarea
purtătorului nu este activ ă pe același UE. Această comparați e este împărțită în două categorii, și
anume pe UE staționare la distanțe de 100 de metri și pentru cazul in care testam efectul schimbări
modului de transmisie asupra debitului .

11

Figur a 2.6 – Viteze de transfer cu și fără agregarea purtătoarelor activă

În imaginea de deasupra sunt prezentate valorile pentru debitul obținut pe UE când agregarea
purtătoarelor este activă versus când nu este activa pe același UE, rezultatele au fost obținute la o
frecvență de 5MHz sau 25 de RB -uri. În imaginea de mai jos sunt prezentate efectele schimbări
modului de transmise a supra debitului .

Figur a 2.7 – Viteze de transfer pe diferite moduri de transmisie

În imaginea de deasupra se poate vedea cum debitul pe dispozitv creste odata ce modul de
tranmsie a fost schimbat de la o singura antenă la două antene sau diversitate in transmisie pentru
o frecvență de 5MHz sau 25 RB -uri . Această creștere a debitului este cel mai vizibilă la distanțe
mari unde se fac prezente efectele atenuări, aceste rezultate fiind obținute la o distanță de 1500 de
metri de eNB folosind modelul de popagare in spațiu liber.
Pentru configura ția in care avem mobilitate l a o viteză de 10 m / s debitul scade constant în
timp ce UE se deplasează tot mai departe de eNB . La o viteză de 30 m / s valorile debitului scad
mult mai rapid decât la viteza de 10 m / s.
Ultimele rezultate se referă la modul în care mecanismele de planifica re MAC se comportă
într-un med iu simulat în ns -3. Rezultatele mecanism ului de ordonare pentru stratul MAC sunt
împărț ite în două părți. Prima parte este pentru mecanisme de ordonare care nu iau în considerare
parametrii QoS, iar a doua parte este p entru cele care iau în considerare QoS.
Topologia în care s -au obținut rezultatele este compus ă din trei configurați i, și anume prim a
configurație in care toate U E-urile sunt poziționate la aceeași distanță de eNB, a do ua configurație

12
în care U E-urile erau poziționate la distanțe diferite, iar a trei a configurație unde UE-urile au viteză
constantă. În imaginea de mai jos este prezentat debitul obținut pentru fiecare planificator MAC
pentru cazul in care UE -urile sunt la aceeași distanță de eNB.

Figur a 2.8 – Debit la aceeași distanță fața de eNB

Chiar dacă toate U E-urile au fost localizate la aceeași distanță de eNB, fiecare algoritm de
ordonare gestionează resursele rețelei într -o manieră diferită. Round robin încearcă să împartă
resursele disponibile în mod egal între U E-uri, valorile pentru proportional fair nu sunt mu lt mai
diferite de robin ro bin, pentru că atenuarea generală pe U E-uri este similară. Maximum throughput
nu este afișat pe acest grafic deoarece atunci când UE -urile au același debit sau prezintă aceleași
caracteristici de propagare, implementarea curentă a algoritmului în ns -3 va selecta întotdeauna
prim ul UE creat în script. În imaginea de mai jos sunt prezentate valorile pentru debit pentru cazul
in care UE -urile sunt poziționate la fel, dar se folosesc algoritmi de ordonare care iau în considerare
QoS, UE -1 si UE -2 au setat un nivel QoS compatibil GBR, iar nivelul QoS pentru UE-3 nu este
GBR .

Figur a 2.9 – Debit pentru UE cu diferit QoS

13
Plasând U E-urile la aceeași distanță elimină majoritatea factorilor care influențează viteza de
transfer, asta permite planificatorilor MAC să ofere rata de transfer maximă fiecărui UE în funcite
de nivelul QoS setat pentru fiecare UE . Round robin deservește toate U E-urile. Priority set
modifica debitul pe fiecare UE in funcție de GBR , iar CQA deservește doar UE -urile ce au setat
GBR , iar cele fără GBR setat pentru ele au alocate resurse doar dacă rămân neutilizate.
Când UE-urile sunt mobile și se mișca din poziții diferite, maximum throughput va maximiza
transferul pe U E-urile care sunt mai aproape de eNB și va ignora pe cele care sunt mai departe
datorită condițiilor de propagare mai rele . In imaginea de mai jos este prezentat valorile pentru
mobilitate p entru maximum throughput.

Figură 2.10 – Variația debitului in funcție de viteză si distantă de eNB

Round robin si proportional fair oferă o performanță similară unul fața de altul pentru cazul
în care sunt mobile UE-urile, deș i proportional fair oferă un debit puțin mai bun și mai stabil ,
deoarece ia în considerare si condițiile de propagare .
Pentru studiul mecanism elor de ordonare a transmiterii pachetelor MA C care iau in
considerare parametrul QoS. Când UE -urile sunt plasate la distanțe diferite există o diferență clară
in modul in care fiecare mecanism de ordinare împarte resursele între dispozitive. Mecanismele de
ordonare a transmiterii pachetelor care țin cont de QoS oferă rezultate complet diferite fața de cele
care nu iau în considerare acest parametru. Pentru cazul în care dispozitivele se mișcă rezultatele
sunt asemănătoare cu cazul in care diapozitivele erau plasate la distanțe diferite.
2.5 Concluzi i
Scopurile urmărite au constat în studiul teoretic al mecanismelor de propagare în rețele LTE
și simularea acestora folosind ns-3. Studiul rețelei LTE efectuat în această lucrare arată că ns -3
poate fi utilizat ca un instrument pentru simularea diferitelor aspecte teoretice LTE și poate fi
generalizat suplimentar pentru a permite studiul procedurilor de transmitere sau semnalizare a
datelor LTE mai complexe. Aspectele acoperite sunt în domeniul transferului de date și a suportului
de mobilitate p e interfața radio și sunt menite să prezinte funcționarea acestor procese într -un mediu
virtual în condiții specifice de mobilitate si modele de propagare . O continuare a acestei lucr ări ar
fi studiul suplimentar al subiec telor deja atinse , folosind topolo gii mai complexe .

14
Work Planning
Task Name Duration Start Finish
Choosing the subject
for diploma thesis 23 days 08.11.2018 30.11.2018
State of the Art and
Theoretical Research 65 days 03.01.2019 19.03.2019
Installation and
configuration of the
ns-3 simulator 36 days 20.03.2019 25.04.2019
Writing of the
theoretical
fundamentals 21 days 03.06.2019 24.06.2019
Creation and testing
of the simulation
scenarios 18 days 25.06.2019 13.07.2019
Writing of the
Implementation 15 days 14.07.2019 28.07.2019
Writing of the
experimental results 15 days 29.07.2019 12.08.2019
Writing the summary
in Romanian and
proof reading 11 days 13.08.2019 23.08.2019
Finishing the work 11 days 27.08.2018 06.09.2019

15
3. State of the art
During this short chapter we will describe the LTE network simulators and the ns-3 platform
together with its features and afferent limitations. We will also offer a short comparison with other
open source platforms for simulating the LTE radio interface.
All these comparisons are made to show the current possibilities in the domain of network
simulation and the importance of this tools in the fi eld of telecommunications.
A simulator main purpose is to build a software simulation model of a networking system. To
analyze, study, improve, develop network protocols. The reasoning behind this is that real systems
are expensive, complex, unavailable and most important there is no infrastructure for the sole
purpose of testing and / or academic purposes.
LTE network simulators are a vital tool for of any engineer or researcher in the field of
telecommunication which makes the study and testing of different configurations or hypotheses
much easier and less time consuming . Some of the fields in which network simulators have made
themselves indispensable are mathematical analysis /modelling of systems, simulation by
modelling the system at abstract le vel via software and emulation of hardware components.
Most LTE network simulators are used for performing active or passive measurements and
experimentation. The advantages of using simulators is that they are relatively easy to set up and
consume less time than trying to perform the same task in real life.
Even with all the advantages that the simulators bring they also come with great disadvantages
that if are not taken into consideration will nullify, they’re usefulness. Some of this potential
disadvantage are: simplified view of complex interactions, possibility of misleading results to the
point of no correlation with real conditions, dependence of assumptions and model.
The ns -3 simulator is a discrete -event network simulator targeted primarily for research and
educational use, [1]. NS-3 is an open source platform, created by the research community, it is not
compatible with the older version or iterations of simulators like ns -2.
The primary use of this software is to provide the means to perform network research and to
be an extensible network simulation platform. Some of the features offered are models for packet
data networks working and performance, simulation engine for variou s experiments, study of the
system behavior in certain scenarios, high rate for obtaining reproductible results over a high
number of simulations.

Some of the defining characteristics of ns -3 are:
• it is designed as a set of libraries that can be combined with other external software libraries
as can be seen in [1] , doesn’t make use of virtualization
• it primarily used on Linux or macOS systems
• it is intended to be used fr om the command line, doesn’t have a native GUI, can use third
party animation software though
• is not an officially supported software product of any company
• can use two scripting languages (C++ and python)

NS-3 has a modular design which allows an easy and fast modifications of the source code to
fit all the needs of the user. Some of the interesting features built in are :
• packets can have “dummy bytes” for simulations where we don’t care about data, no
memory is allocated to this type of bytes and as a result reduces the memory footprint of
the simulation

16
• nodes have optional features like: no memory waste for IPv4 stack for nodes that don’t use
IPv4 protocol, the mobility model is optional (can be removed for certain nodes)
• some of the cross -layer features are: packet tags, tracing
• real world integration features: real -time scheduler, network simulation cradle, packet scan
be saved as PCAP files

Other open source alternatives for simulating the LTE radio interface are funda mental different
from the ns -3 platform since they are more orientated towards simulating computer networks and
don’t feature a detailed simulation for LTE radio interface .
Some of the open source alternative are: Cloonix, Common Open Research Emulator (CORE),
GNS3 , IMUNES, SimuLTE, Mininet, Netkit, VNX and VNUML, etc.
Most of the platforms listed above offer simulating environments for computer networks, they
can simulate LTE network but not in detail since there are paid software’s that offer better support.
One exception is SimuLTE, who is purposely built to simulate LTE radio networks.
From the p erspective of software design , most open source simulators from those listed above
use virtualization and come with graphic user interface that takes care of all the user input,
command line is only used as an option and it’s not intended to be the main method of interaction
with the software as in the case of ns-3.

Another differentiation between ns -3 and its competitors is the fact that ns -3 requires its users
to have coding skill s, since the environment wasn’t made with a user interface in mind.
A common characteristic of all this open source platforms is that they’re based on Linux or
macOS, support for windows its limited and doesn’t have the same amount of traction from the
developers as for Linux and macOS. When it comes to the paid alternatives there are a few of them.
Some of them are: Qualnet, NetSim, OmNet, OpNet . A short description for each of them is
provided belo w.

OmNeT is a C++ based discrete event simulator for modeling communication networks and
other distributed or parallel systems. [2]
OPNET Network simulator is a tool to simulate the behavior and performance of any type of
network. The main difference O PNET Network Simula tor comparing to other simulators lies in its
power and versatility. It provides pre -built models of protocols and devices. It allo ws to create and
simulation different network topologies. [3]
NetSim is an end -to-end, full stack, packet level network simulator and emulator. It provides
network engineers with a technology development environment for protocol modeling, network
R&D and military communications. [4]
QualNet network simulation software is a planning, testing, and training tool that "mimics" t he
behavior of a real communications network,[5] . Qualnet provides a n environment for designing
protocols, creating and animating network scenarios, and analyzing their performance.

The proprietary software is more feature rich than the open source counte rpart and more user
friendly since the scenarios set up is done from the user interface. Most of the tools listed above
offer academic licenses to universities, making them easier to obtain for students.
A bonus of using licensed software is the support offered, updates for bugs and new features,
this type of support is also present on the open source alternatives, but it is not guaranteed , meaning
that the user will have to get creative or improvise as it goes.

17
When it comes to the operating system on which the application is based, the predominant one
is windows, this is the case because most business and universities use Microsoft windows
operating system.
The LTE module developed for ns -3 focuses mainly on modeling the E-UTRA (Evolved
UMTS Terrestrial Radio Access ) part of the system. The module has basic implementation for LTE
devices, including propagation models for LTE Radio Protocol stack (RRC, PDCP, RLC, MAC,
PHY).
NS-3 can simulate also the EPC model. This model i ncludes network interfaces, protocols and
entities, which reside within SGW, PGW nodes, and partially within the eNB. The EPC model
objective is to provide means for simulation of IP connectivity over the LTE model .

The LTE model has been designed to simu late a variety of aspects of the LTE network some of the
are:
• Radio Resource Management
• QoS-aware Packet Scheduling
• Inter -cell Interference Coordination
• Dynamic Spectrum Access

Besides simulating the components of the network, ns3 can also simulate to different degrees
of realism other aspects of true telecommunication network. Some of these aspects are channel
propagation model, fading model , MIMO model , MAC scheduling, QCI leve ls, RLC and so on.

18
4. Theoretical fundamentals
4.1 Introduction
The phenomenal success of GSM built the foundation of the circuit switch ing system . During
1990s as the popularity of the internet services grew more and more it become apparent the need
for mobile internet. The first mobile internet services were very limited , both because of the limited
processing capability of the early mobile devices and the limitations of bandwidth for the radio
interface. [6]
Later this status quo changed as the first big push in the evolution of the radio access network
took the place . This revolutionary architecture was called UMTS, but even though it made the
mobile internet more accessible and reliable , it was still not enough. As the technology matured it
laid the foundations for the next revolution in broadband radio access and rapid convergence of
internet and mobile services, this next step was called LTE.
LTE was designed and build to respond to the growing nee ds of the new emerging technologies
such as multimedia applications , online gaming, mobile TV, Web 2.0, streaming contents. After
taking these requirements into consideration it follows that the goal of LTE was and is to provide
high data rate , low latency , packet optimizing and a support for flexible bandwidth deployments.
While the term “LTE” encompasses the evolution of the UMTS from the UTRAN to E –
UTRAN. EPS covers the evolution of other aspects, commonly known by the term of SAE , one
shining example is the EPC. LTE and SAE together manage to form the EP C also called the core
network.
The core network is the part that bridges two different worlds. On one side we have the high –
speed radio access technologies and on the other hand we have the fast -changing environment
characteristic to the place we call the internet. The evolution of the core network is an important
achievement and the corner stone of the modern mobile broadband revolution. Without this
fundamental component neither RAN nor mobile internet services would manage to achieve their
full potential.
While the evolution of the core network offered great features, the radio access network also
made use extensively of concepts like FD D, where the uplink and downlink channel s are separated
in frequency, and TDD, where they share the same frequency channel but are separated in time.
These concepts materialized into the modern air interface composed out of frequency division
multiple access for downlink and uplink transmission to achieve high peak data rates in
environments with both a high amount of bandwidth available and in the ones where bandwidth
comes at a premium price.
4.2 LTE network architecture overview
As previously stated, EPS evolved from its predecessor UMTS, in doing so it kept some
similarities with its predecessor, but also deviated from UMTS in some key areas. This key area
being the migrati on from CS to PS and up to an all-IP PS -only network . Because of this change the
LTE architecture become a flat architecture where the user plane tunneled directly between the
core and access networks. This change of architecture can be seen in the image below.

19

Figure 4.1 – Comparison between 3G and EPS architecture adapted from [7]

The new architecture was set to bring many features and advantages over its predecessors at
the cost of setting high level requirements for the new technology . Some of these requirements are:
• High er transfer speeds : High er data rates can be achieved in both downlink and uplink
• Low latency : Time required to connect to the network is very small , small IP packets can
have a latency of under 10ms for services like VoIP, gaming, streaming
• Low power consumption: now it is easier for the user device to enter and exit different
power stages, this is made possible thanks to the eNB (E volved Node B)
• FDD and TDD used together : FDD and TDD being used in tandem to ensure a higher
efficien t use of the limited resources
• Superior end-user experience : optimized and more seamless connection of the user to the
network. Another factor is the decrease of overhead signaling
• Simple architecture: because LTE has only one core systems that handles all types of traffic
and signaling, the cost of using and maintaining processes is small in terms of resource s

In terms of actual figures that define the system and its capabilities t he initial releases and
standards envisioned speed of 100 Mbps for download and up to 50 Mbps for upload for every 20
MHz of spectrum used. Furthermore, LTE was required to serve up to 200 active users in every 5
MHz cell.
4.3 Detailed breakdown of EPS architecture
In the previous chapter we talked about EPS architecture and made a short comparation to
UMTS to put in context the differences between the two architectures . In this chapter we will walk
through the individual components that encompass EPS network . A more detailed view of LTE
architecture is shown in the image below.

Figure 4.2 – EPS block scheme architecture adapted from [8]

20
EPS is a term that refers to the complete end -to-end system, that is UE, E -UTRAN and core
network. EPS is a system that works with IP connectivity and a PDN to provide the user with access
to internet. This structure is used for different type of services , an example of this serv ices is VoIP .
A characteristic of this system is the EPS bearer , it is typically associated with QoS. The bearer
works like a resource that is used to provide different level of QoS depending on the connectivity
to different PDNs. A classic example is described in [8] where a user who is engaged in voice call,
VoIP, while at the same time performing other tasks like web browsing, downloading large
documents or streaming. To satisfy the needs of different levels of QoS for each application, we
need different QoS streams for each distinct flow of data. For instance, a VoIP bearer would provide
the required QoS for the voice call , while for the web browsing a best effort bearer would be suited
to provide the required level of QoS.
To achieve this performance several EPS components that have different roles work together,
each doing a specific operation . Though each component performs a different task, they are
organized in distinct functional blocks. This can be viewed as a form of modularity which helps in
managing all t he functionalities required for managing the network.
To be more precise the EPS architecture is comprised from the following three main
components:
• The user equipment.
• The evolved UMTS terrestrial radio access network.
• The evolved packet core also referred to as core network

Each component serves a broader purpose that its first let to believe. Next, we will take each
one of the three components and present their role and place in the bigger picture.
4.3.1 Evolved packet core
EPC contains most of the infrastructure that deals with the authentication, mobility
management, traffic routing, self -diagnosis and maintenance. The evolved packet core is mainly
composed out of MME , S-GW and P-GW. In addition to this main building blocks EPC
encompasses other logical nodes, which come in play to help the different systems mentioned
above. These logical nodes are: HSS, PCRF.
HSS, the home subscriber server contains all the subscriber data . The information contained is
both used for authentication and to manage the resource allocation. An example of this data is the
QoS profile and any restrictions in roaming.
In [8] PCRF is tasked with the control and decision -making for QoS authorization , whic h
decides how a certain data flow will be handled by the PCEF . PCEF resides in PC RF and ensures
that the subscriber is being billed and allow ed to use the resources in accordance with its
subscription.
A general definition given in [8] des cribes P-GW as the PDN Gateway which deals with IP
address allocation for UEs, QoS enforcement and charging according with the policy of the PCRF.
Another task performed by the P -GW is the filtering of IPs packets on the downlink direction
according to the used QoS b earers, this is performed according to predefined TTFs. QoS
enforcement is used for ensuring GBR rate for each user according with its subscription and policy
of PCRF.
S-GW, the serv ing gateway as the name implies is a gateway for all users IP packets, whi ch
need to be transferred. S -GW retains some information about each UE which services, such
information is related to which bearers the UE uses, this information is mainly used after the UE
performs a handover between eNB s or when the device goes into idle state. The responsibilities of
S-GW mainly focus on: handovers , mobility interface , monitoring and maintaining context
information related to UE .

21
[8] describes MME as the mobility management entity is a special node whose sole job is to
process signaling information between UE and CN, this signaling information comes mainly from
the NAS protocols . Main function s performed by MME can be classified as:
• Bearer management – includes the establishment, maintenance and release of the bearers
• Connection management – includes the establishment of the connection , security between
the EPS and UE , it is handled by the connection or mobility management layer
4.3.2 Evolved universal terrestrial radio access network (E -UTRAN)
E-UTRAN can be said to be the essential component of the EPS structure, we say this because
it performs only one task , that of connecting the UE with the rest of the core network. E -UTRAN
as the rest of the LTE network shares some of its characteristics with UMTS, this similarity can be
observed in the naming scheme. In a nutshell E -UTRAN is composed out of a network of eNB s.
An example of such network can be viewed in the image below where X2 stands for X2 interface.

Figure 4.3 – E -UTRAN block scheme

One eNB typically consists of three antennas and its usually connected with other eNB via the
X2 interface, the interface has functions for mobility and load exchange. Since the main goal of
eNB is to connect with the UE, it has a specialized interface called LTE -Uu or air interface . This
interface is b ased on OFDM technique , namely OFDMA in downlink and SC -FDMA in uplink.
[8] defines E-UTRAN as being responsible for all radio -related functions, which can be
summarized as:
• RRM – it performs all functions related to radio bearers, such as radio bearer control, radio
admission control, radio mobility control, scheduling and dynamic allocation of resources
to UEs in both uplink and downlink.
• Header Compression – mainly deals in comp ressing IP packets headers, this i s done to
reduce the overhead in the network and to increase the efficient use of the radio interface
• Security – All data sent over the radio interface is encrypted.
• Connectivity to the EPC – consists of the signaling toward MME and the bearer path toward
the S -GW.
4.3.3 The user equipment
The user equipment or mobile device is the part of the LTE architecture that most people
interact on a day to day basis, from the point of view of an average user this is the only aspect of
the network that they know about and that hasn’t change d despite the constant evolution of the
architectures.
From a technical point of view the user equipment is seen as a part of the network that has not
changed a lot form the days of UMTS and GSM. This is said because the internal structure of the

22
user equip ment for LTE is almost identical to the one in UMTS and GSM and which is ME (Mobile
Equipment).
The ME components are:
• MT: handles all the communication functions.
• TE: terminates the data streams.
• UICC: also known as the SIM card for LTE. Runs an application known as the USIM .
4.4 LTE physical layer
No discussion about LTE is complete without taking the time to talk about the influence and
impact of the physical layer on the LTE network . Physical layer had a major influence on the design
and functioning of the LTE network , starting with radio access technique, continuing with the way
the frequency reuse is performed and how techniques like MIMO are employed in LTE.
4.4.1 OFDMA and SC -FDMA
OFDM is the basic signal format used in LTE , although OFDM is not used in its original form,
its variants are being use d in the same way , meaning multiple access scheme. OFDMA in downlink
and SC -FDMA on the uplink .[9]
OFDM and its variants offer great resilience against narrow band fading, this impressive
property of OFDM is because it uses multiple narrowband carriers, each carrier carrying a low data
rate. The LTE OFDM signal is usually made from modulations like PSK and QAM, higher order
modulation can be used to achieve higher data rates. [9]
The implementation of OFDM technique is different for downlink and uplink, this was
necessary to meet the different requirements faced in the two opposite ends of the air interface. The
main factor at play in using different implementations for UE and eNB is the equipment power
requirements. Another important factor in choosing OFDM is that it can be used both in FDD and
TDD and that this modulation is very suitable for high data rates. [9]
One of the key factors taking into considerations for the LTE OFDM are the choice of
bandwidth. The amount of bandwidth and its location in the spectrum influenc ed a variety of
decisions like the number of carriers, which in turn influences the symbol length. A set of
bandwidths has been selected for LTE and it is : 1.4 MH z, 3 MH z, 5 MHz, 10 MHz, 15 MHz, 20
MHz
For the downlink part it is used OFDMA . Each user is allocated a set of resource blocks , a
resource block is composed out of 12 carriers and 6 or 7 symbols on each carrier . The resource
blocks translate into channel bandwidth which is the same with the set of bandwidths selected for
LTE. The more resource blocks a user has the higher the transfer rate, the transfer rate or throughput
is related also to the type of modulation used for the allocated resources. The allocation and the
task of managing t hem belongs to the scheduling mechanism. For modulation is possible to choose
between three types of modulation: QPSK/4QAM, 16 QAM, 64 QAM and 128 QAM for LTE -A.
For the uplink part it is used SC-FDMA . The decision of using this technique was motivated
by the key characteristic that affects every mobile device , battery life. Even though battery life
improves each generation, it is still necessary to ensure a battery consumption as low as possible.
Even if it doesn’t seem like this at first , but the radio frequency amplifier inside the mobile device
is one of the highest power -hungry items , by using SC-FDMA for uplink we manage to maintain
the power consumption to minimum.

23

Figure 4.4 – Downlink resource grid for 1.4MHz
4.4.2 Pathloss models
An important part of the physical layer in LTE is path loss or path attenuation. Since LTE is a
mobile network which relies on the air interface, we need to put into context the dependency of the
network with the path attenuation or path loss by using so called propagation models.
Path loss is in broad terms a reduction of the power density of an electromagnetic wave as it
propagates through the s urrounding environment, which can be air, vacuum and even solid objects
like buildings walls or surrounding plants. By being such a widespread phenomenon, it follows that
it plays a major role in the planning and design of the link budget of a telecommunic ation system.
Now that one has a better understanding of the path loss phenomena, we can look at the
propagation models.
Friis propagation model, this model is only valid for propagation in free space in a LOS (Line
of Sight) settings where there are no ob jects to create absorption, diffraction or reflections. This
model is based on the inverse square law of distance which is presented below.

𝑃𝑟
𝑃𝑡=𝑃𝑡𝐺𝑡𝐺𝑟𝜆2
(4𝜋𝑑)2𝐿 (1)

The two-ray ground propagation loss model, this model predicts the losses which occur
between two antennas situated at different heights in relation to each other and are in LOS. The
model describes a signal with two components, one LOS component and a multipath component
formed by LOS component after being reflected by the ground. The formula for calculus is:

𝑃𝑟=𝑃𝑡𝐺𝑡𝐺𝑟𝐻𝑡2𝐻𝑟2
𝑑4𝐿 (2)

Another path loss model, valid also for LTE simulation is Okumura Hata propagation model,
it implements an open area pathloss for long distances over on e kilometer by using the COST321
equation, presented below:

24
𝐿=46.3+39log𝑓−13.82logℎ𝑏+(44.9−6.55logℎ𝑏)log𝑑−𝐹(ℎ𝑀)+𝐶 (3)

Additionally, other path loss models like three log distance propagation loss model is a viable
option for LTE simulations. This propagation model computes the log distance path loss for three
distances: near, middle and far with different exponents. The formula for calculus is:

𝐿 =
{ 0 ,𝑑<𝑑0
𝐿0+10log10(𝑑
𝑑0) ,𝑑0≤𝑑<𝑑1
𝐿0+10∙𝑛0log10(𝑑1
𝑑0)+10∙𝑛1log10(𝑑
𝑑1) , 𝑑1≤𝑑<𝑑2
𝐿0+10∙𝑛0log10(𝑑1
𝑑0)+10∙𝑛1log10(𝑑2
𝑑1)+10∙𝑛2log10(𝑑
𝑑2) , 𝑑2<𝑑

(4)

More details about each pathloss model described above can be found in [13].
4.4.3 Frequency reuse algorithm s
As in any other wireless communication network the scarcity of bandwidth, the limited radio
spectrum and the power limitations has made the need for an efficient way of using available
bandwidth more necessary . A solution to this problem is the reuse of ex isting frequencies in
adjacent cell s or in a pattern, this solution is not a perfect one as it comes with some disadvantages,
mainly the ICI. Further research and development have brought this concept into every wireless
network including LTE. In LTE ther e are present two types of frequency reuse algorithms: SFFR
and SFR.
SFFR is an alteration of the traditional frequency reuse scheme used. Each cell has two zones
of frequency allocation, a central one that is common to all or multiple cell and the zone around
the edges of the cell who employs another frequency than the center part of the cell which is
different for each cell. This edg e frequencies are allocated in accordance with an algorithm to keep
ICI to minimum between cell belonging to a group.
SFR algorithm employs the same treatment for edge frequencies as SFFR, but the interior user s
can use the sub -bands usually reserved for the edge users. Because cell -interior users can share the
bandwidth with neighboring cells they have to transmit at a lower power level than the cell -edge
users. SFR is more efficient than SFFR in terms of effici ent use of bandwidth, but it results in more
ICI for both center and edge users. More details on the theme of frequency reuse for both algorithms
can be fo und in [ 8].
4.4.4 MIMO technique s
The basic concept behind MIMO is multipath signal propagation. In other words, it utilizes
more than one antenna at a time to offer increased performance. The transmitter and receiver have
more than one antenna and are able to use the different paths that are present between them to
improve the SNR. Currently in LTE are several transmission modes which can be employed while
using MIMO, some of them are:
• Single antenna, this is the most basic form of wireless transmission. A single antenna
transmitting a si ngle data stream and is received by one or multiple antennas.
• Transmit diversity: it transmits a stream of information from multiple antennas. LTE
support two or four antennas for this mode.
• Open loop spatial multiplexing: this form of MIMO sends two information streams which
are transmitted over two or more antennas. There is no feedback from UE.

25
• Close loop spatial multiplexing: is like open loop version but has feedback mechanism
integrated to close the loop.

A more detailed explanations of each mo del and the additionally available transmission modes
in LTE can be found in [7].
4.5 LTE radio protocol architecture
The radio protocol architecture for LTE can be divided into two separated layers , this being
the control plane and user plane. On the user plane side, the application creates data packets which
are sent by protocols such as TCP, UDP and IP. On the others s ide the control plane has the RRC
protocol which writes the signaling messages that are exchanged between eNB and UE. For both
planes the information is processed by PDCP, RLC and MAC protocols, before being passed to the
physical layer.
4.5.1 User plane
In [8] the user plane describes the packet s between UE and EPC as being encapsulated in a
specific protocol and after this is tunneled between the P -GW and eNB for an eventual transmission
to UE. There is no default tunneling protocol, because it varies ac ross different interfaces. An
example of 3GPP tunneling protocol is GTP is used between eNB and S -GW and between S -GW
and P -GW.

The user plane protocol stack between the eNB and UE consists of:
• Packet data convergence protocol
• Radio link control
• Medium access control

Packet s that are received by a layer are called SDU , while the packet output of a layer is called
PDU, the IP packets of a user flow from top to bottom layers.
The figure below shows the protocol stack for the user -plane, the same protocols are presen t
in both sides of the air interface . Some of the functio ns performed by the user plane are header
compression, ciphering, scheduling mechanism and more others which can be found in [8] .

Figure 4.5 – User Plane Protocol Stack adapted from [8]

26
4.5.2 Control plane
The control plane has other functions additionally to the ones provided by RRC, which is
responsible for configuring the lower layers. This additionally functions are for handling radio –
specific functionalities which depend on the state of the UE, the states being idle or connected.
The protocol stack shown in the figure below , is for the connection between UE and MME.
Lower layers have the same functions as in the user plane with the added benefit of not needing to
compress the header.

Figure 4.6 – Control plane protocol stack adapted from [8]

The physical layer mainly carries information for MAC transport channels. Furthermore, it
performs a series of administrative functions like link adaptation, power control, cell search and
different types of measurements for RRC.
The MAC layer is responsible for managing and mapping packets between logical channels
and transport channels, multiplexing and d e-multiplexing SDU packets fr om and to the physical
layer on transport channels. In addition, it performs scheduling information reporting, error
correction through HARQ, priority handling by dynamic scheduling for UEs and logical channel
prioritization. More details on the working of MAC layer can be found in [7].
The RLC operates in 3 modes, each mode being responsible for implementing one or several
functions , being TM, UM and AM. The functions and which mode is responsible for are:
• error correction through ARQ , only for AM data transfer
• Concatenation , available for all modes of operation
• segmentation and reassembly of RLC SDUs , only for UM and AM data transfer
• re-segmentation of RLC data PDUs , only for AM data transfer
• reordering of RLC data PDUs , only for UM and AM data transfer
• duplicate detection , only for UM and AM data transfer
• RLC SDU discard , only for UM and AM data transfer
• RLC re -establishment and protocol error detection, only for AM data transfer

The RRC main function is the broadcast of system information to different protocols and a
slew of other services, which are: broadcast of information related to NAS and AS, paging ,
maintenance, e stablishment and release of an RRC connection between the UE and E -UTRAN ,
key management for security , configuration and release of point to point radio bearers.
PDCP is an important protocol that manages t he establish ment of the connection between the
user and control plane. The protocol sits at the center of the network and is responsible for many
functions , some of them are: header compression and decompression of IP packets , transfer of data

27
in user plane or control plane , maintenance of PDCP sequence number , ciphering and deciphering
of control plane or user plane data.
4.6 Quality of Service
[7] define s QoS as the ability to support a variety of non -real-time and real -time applications
at the same time, in a manner such as to not become a reliability to the network . Packet switched
networks like LTE can classify, schedule and forward traffic based on the destination address and
the type of media being used.
The QoS in LTE is done with the help of radio bearers, which are provided to the eNB by the
MME using the standardized QoS attributes. Based on thes e attributes the EPS and protocol layers
between UE and eNB can manage scheduling of uplink and downlink traffic. [7]
Mainly , bearers can be classified into two categories, depending on the nature of QoS they
provide:
• Minimum guaranteed bit rate bearers which are used for application like VoIP. These have
set a bit rate value for which resources are permanently allocated by the netwo rk at bearer
establishmen t.
• Non-GBR bearers, these bearers do not guarantee any particular bit rate. These are used for
applications like FTP transfer and web browsing. For this class of bearers, no resources are
allocated permanently.

eNBs are responsible to ensure the necessary QoS for a bearer over the radio interface.
Additiona lly, each bearer has an associated QCI and an ARP . QCI is characterized by priority,
packet delay budget and acceptable packet loss rate. The QCI label for a bearer determines the
importance of the bearer inside eNB. Some QCI have been standardized so that everybody can
have the same understanding of the service .
The priority and packet delay budget f rom QCI determine how the RLC makes its
configuration and how the M AC scheduler handles packets sent on the bearer. For example, a
packet with a higher priority is scheduled before a packet with lower priority. Some standardized
QCI are presented in the table below.

Table 1. Possible QoS levels adapted f rom [7]
QCI Resource
types Priority Packet delay
budget (ms) Packet Error
Loss rate Example Services
1 GBR 2 100 10−2 Conversational Voice
2 GBR 4 150 10−3 Conversational Video (live
streaming)
3 GBR 5 300 10−6 Non – Conversational Video
(buffering video)
4 GBR 3 50 10−3 Real Time Gaming
5 NON -GBR 1 100 10−6 IMS Signaling
6 NON -GBR 7 100 10−3 Voice, Video, Gaming
7 NON -GBR 6 300 10−6 Video (buffered gaming)
8 NON -GBR 8 300 10−6 TCP-based chat, FTP, video
9 NON -GBR 9 300 10−6 TCP-based chat, FTP, video

IP packets mapped on the same EPS bearer receive the same treatment. Providing different
bearer -level QoS requires the EPS to filter the IP packets on different bearers which have assigned
other levels of QoS. Packet filtering is based on TFT for different bearers . The TFT uses the
information present in the header of the IP packet so that each packet is sent to its corresponding
bearer with the appropriate level of QoS. TFT filters are present for both uplink and downlink.

28
When a UE attaches to the network as part of the procedure it receives an IP address from P –
GW and a bearer with a default QoS level, this bearer remains with the UE on the entire lifetime
of the PDN connection so that the UE will have always -on IP connectivity. The initial bearer QoS
level is given by the MME, based on the subscriber information present in HSS. The PCEF may
change this value in interaction with PCRF or if it receives a request f rom the UE or other entity.
Additional other bearers can be set, which can be GBR or non -GBR.
4.6.1 MAC schedulers
The MAC protocol exists in UE and eNB and it is a vital part in the correct functioning of
these two entities in relation to each other and with the rest of the network. The role of the MAC
layer ca be best observed over the air interface on the user and control planes and controls the
access to the shared transmission medium.
The MAC scheduler manages a structure called DCI, this structure is then transmitted by eNB
to the connected UEs. This is done to inform the UE with regards to the resource allocation on a
per sub -frame basis for downlink direction. MAC scheduler populates the DCI structure with
specific information, such as: MCS to be used, the MAC transport block size and the bitmap
allocation which classifies which RB will contain the data transmitted to each user by eNB.
The procedure described above is based on the CQI, this measurement unit is used to quantize
the spectral efficiency and is rounded to the lowest values. This scheme adapts the MCS selection
to the PHY layer performance according to the CQI report.
MCS is related to the modulation order used by the PHY layer t o decide what type of
modulation to use to transmit information. MCS is used in both uplink and downlink to determine
the type of modulation used. Depending on the MCS the transport block size is decided . The values
for MCS vary from 0 to 28, where 0 represents the lowest modulation order , represented by QPSK
modulation, and 28 is among the highest order of modulation, usually represented by 64 QAM
modulation. More details on the role and importance of MCS can be found in [7].
Among the various packet scheduling algorithm available in the literature ns -3 offers support
for only a limited set which is described in [10] , some of them are shortly reviewed below.

Round robin scheduler is the simplest scheduler. It works by dividing the available resources
among all the active connections and flows regardless of priority or QoS. The MCS given to each
user is done according to the received CQI. [10]
Proportional fair scheduler is more complex than round robin and works by scheduling based
on its own average channel condition, when a user has an instantaneous channel quality higher than
the average conditions it receives more RB regardless of the QoS setting .[10]
Maximum throughput scheduler is different than round robin and proportional fair by the fact
that it aims to maximize overall throughput of the eNB. It does this maximization by allocating
RBGs to the user who can achieve the maximum throughput in the current transmission ti me
interval based on the reported CQI indicator . [10]
Priority set scheduler is a QoS aware scheduler which combines frequency domain and time
domain packet scheduling operations in one scheduler. It controls the scheduling among UEs by
using TBR. [10]
CQA (Channel and QoS Aware) scheduler is a MAC scheduler used on the downlink side, the
algorithm used for scheduling works with HOL (Head of Line), the GBR parameter and channel
quality. It groups user by priority to enforce the scheduling and prioritizes the flows of information
to the users with the highest priority. [10]

29
4.7 Carrier Aggregation
To fully understand the concept and goal of carrier aggregation we first need to take a look at
proposed feature set for LTE advanced. LTE advanced is a major enhancement of the traditional
LTE, which was designed to support a maximum bandwidth of 100Mhz a nd peak data rates of 1
Gbps in the downlink and 500 Mbps in the uplink. This requirement of extremely large bandwidth
is very unlikely to be available in most cases and it became the motivation for the implementation
of the concept called carrier aggregation.
Another motivation for carrier aggregation is to make more efficient use of the fragmented
spectrum. Carrier aggregation is designed to use multiple frequency blocks, called CC which are
assigned to a user .
CA present in LTE advanced is designed to support aggregation of a variety of different
arrangements of CCs and load balancing. These arrangements of CCs can be of the same or
different frequency bands.
Carrier aggregation affects the PHY layer, RRC, MAC protocols, S 1-AP and X2 -AP signaling
protocols on the air interface , there is no impact on RLC or PDCP protocols , CA can be used for
both FDD and TDD. An important note to keep in mind when discussing about CA is the fact that
is possible to use up to 5 carriers at the same time, with the condition that the maximum aggregated
bandwidth is 100 MHz. Usually the number CC for uplink and downlink is different . The easiest
way and more common way of using CA is to use contiguous allocated CCs within the same
operating frequency, this mode is called intra -band. Non-contiguous allocation is also possible if
the spectrum is fragmented.
In the downlink direction the component carriers are classified into two categories, namely
PCC (Primary Component Carrier) and SCC (Secondary Component Carrier). PCC deals with
information like security parameters, NAS informatio n, RRC and other signaling information, the
PCC changes only at handover . SCC is composed out of the rest of component carriers; these
component carriers deal mainly with the transmission of data and ca n be added and removed easily
and as often as required.
4.8 Handover procedure
Handover in LTE plays a major role in its functioning as a network. Like i n other networks
which came before LTE the problem of mobility presents big challenges, more so than ever becaus e
LTE is the first network designed to be an all IP mobile network. Handover presents a high level
of complexity in LTE since the network must deal with multiple types of handover between
different entities in the network.
In LTE , only hard handover is supported, t his being said the main goals of the handover
procedure in LTE are maintaining of the QoS level, keeping the level of power consumption at a
minimum so that the battery of the user device will not be drained, the UE should be able to
continue using the services before and after the handover procedure and also being able t o perform
handoff between other networks and technologies like 2G, 3G.
To do all the task presented above LTE has implemented handover algorithms based on events
in the network. These events are used to define different types of handover algorithms based o n
measurements reports from UE. The measurements are mainly performed on the downlink channel
and are based on the reference symbols, namely RSRP and RSRQ.
The events who make use of RSRP and RSRQ are A1, A2, A3, A4, A5, each event has a
different interpretation for RSRP and RSRQ, namely:
• A1 is triggered when the serving cell becomes better than a given threshold
• A2 is triggered when the serving cell becomes worse than a given threshold
• A3 is triggered when the neighboring cell is better than the serving cell by a given offset
• A4 is triggered when the neighboring cell is better than a given threshold

30
• A5 is triggered when the serving cell becomes worse than a given threshold while the
neighboring than the same given threshold, this e vent needs both conditions to be true

More information about the handover events is present in [8]. Although LTE network deals
with a big variety of handovers, the most common one is the X2 based handover. The procedure
for X2 handover is not so complicated, because the eNB can decide by itself how to hand the mobile
to another cell.
RSRP is a measurement of type RSSI, it is defined as the linear average over the power
contributions of the elements that carry specific cell reference signal s within the used frequency
bandwidth. RSRP measurement is usually expressed in d Bm, it is utilized to evaluate signal
strength among different cells . RSRQ provides additional information and is usually used make
handovers and cell selection decision . The RSRQ is defined as the ratio as the ratio between RSRP
and RSSI, it depends on the amount of bandwidth used and the number of RBs. RSSI is used to
determine the interference and noise information.
Another important parameter for this whole process is SINR . SINR is defined as the ratio of
the average received power to the sum of the average co -channel interference and other noise
sources power. In the image below is shown how a typical X2 handover mechanism takes place in
an LTE network.

Figure 4.7 – Handover procedure adapted f rom [8]

More information about the handover procedure and other aspects of the handover process can
be found in [6], [7] and [8] with different degrees of details regarding each subject.

31
4.9 Link adaptation
As in any wireless communication network the LTE network faces the same problems when it
comes to the issue of signal quality and system capacity optimization of the services offered by the
network to each UE. T he quality of the signal depends on the quality of the communication channel
and factors which influence it, these factors being the type and influence of the propagation loss or
attenuation present at that gi ven moment , the state of the UE , stationary or mobil e, multipath fading
and the interference and noise present on radio link.
To face all these problems the LTE network uses a mechanism called link adaptation designed
to optimize the system capacity and coverage at a given transmission power by matching the
amount of information or data rate sent to each individual user to the signal quality reported by
each user. The process performed by link adaptation is commonly based on the adaptive modulation
and codin g technique .
Adaptive modulation an d coding is a channel aware technique that works by adapting the MCS
to the reported channel quality . The measurement unit used in this process by the adaptive
modulation and coding to modify the MCS is CQI and is reported by each user to the network. The
CQI is typically obtained from the measurement of downlink signal quality by the UE and is not a
direct indication of SINR.
AMC woks by modifying two aspects of the radio interface connection between UE and eNB.
These two aspects are modulation scheme and coding rate. Modulation scheme works by adapting
the type of modulation used in the transmission of information to the characteristics of the channel,
it does this by using low order modulations for high levels of interference on the channel and high
order modulation for low levels of interference and noise on the channel to obtain high bit rate s.
Code rate adaptation works by choosing the best suited code rate depending on radio link
conditions . The process of choosing a code rate is similar to that used in choosing the modulation
scheme, low er code rate s are used for poor channel quality radio link s and high code rates are used
for high quality radio link s with high SINR.
The possible values for modulation scheme and code rate as a function of CQI are displayed
in the table below .

Table 2. Mapping between CQI and modulation scheme and coding scheme adapted from [8]
CQI index Modulation scheme Approximate code rate
0 Out of range —
1 QPSK 0.076
2 QPSK 0.12
3 QPSK 0.19
4 QPSK 0.3
5 QPSK 0.44
6 QPSK 0.59
7 16QAM 0.37
8 16QAM 0.48
9 16QAM 0.6
10 64QAM 0.45
11 64QAM 0.55
12 64QAM 0.65
13 64QAM 0.75
14 64QAM 0.85
15 64QAM 0.93

32
5. Implementation
5.1 Introduction to NS -3
NS-3 is an open source discrete -event network simulator designed for research and educational
use. NS -3 was developed to provide an open, flexible and extensible simulation tool which provides
a simulation engine for user to conduct experiments, testing, studies of systems behavior in
controlled environment or to simply learn about the inner working of networks being internet based
one or not.
By being an open source software ns -3 has the freedom of being more community friendly
than its licensed alternatives. This freedom translates in the fact that ns -3 is designed as set of
libraries which are available to eve rybody who desires to work with or contribute to ns -3. By being
designed as a set of libraries ns -3 retains a level of modularity which enable the user to use other
software to analyze, visualize and compute the data.
The modularity of ns -3 also allow s the software to use two scripting languages C++ and
Python, this is convenient. In terms of supported operating systems ns -3 is intended to be primarily
used on Linux or macOS systems, support for Windows operating system exist and it requires
visual stu dio to be installed on the machine.
Before heading to the part of showcasing the structure of a ns -3 script and the work done in
ns-3 we first need to establish some ground rules and concepts called key abstractions, we do this
to avoid potential misunders tandings and unwanted comparisons with the real world or with what
is already consecrated in theory.
Since ns -3 is not solely intended to work as internet -based network simulator we shall name
any computing device, or entity that we wish to simulate with a general term, this being the node,
name which originates from the graph theory.
Next on the list is the application. Onc e one has created the nodes next step is to install an
application on them to make them useful. We run the application on the nodes to drive the simulated
world.
Similarly, to the real world we need to use a channel in the same way we connect a computer
to internet by using an ethernet cable. In the virtual world of ns -3 we connect a node to
communication channel which can emulate a wide variety of environments from a normal wire to
a full working radio access network.
Net device is an abstraction that covers both simulated hard ware and software driver. The role
of the net device is similarly to the network card in a computer, by installing a net deice to a node
we enable the node the capability to communicate with another node on the previous created
channel. Furthermore, by ins talling multiple net devices to a node we can connect the node to
multiple communication channels.
Topology helpers are different from what w as described about until know in the way that they
do not emulate something specific, at least not directly. Topolo gy helpers are viewed as tools to
arrange the connections between nodes, channels and net devices, they work as the name implies
as helpers and makes the very common work of connecting and setting parameters for very node
easy and less time consuming.
Next in the introduction into ns -3 is presented the structure of a basic simulation script of a
simple LTE network. The picture on the next page puts into view the minimal required simulation
program structure that is needed to do an LTE – only simulation, wi thout any EPC entities.

33

Figure 5.1 – NS-3 LTE simulation general workflow

To offer a better understanding of the structure presented in the above image we will go
through each step and show the commands and offer a short description.
So, the initial boiler plate is a line of code which configures the emacs mode line. This informs
the emacs on which kind of formatting conventions we use in our code.

“/* -*- Mode: C++; c -file-style: "gnu"; indent -tabs-mode: nil; -*- */”

Emacs (Editor MACroS) is a family of text editors used primary to write code, emacs are a
platform which includes built -in command s that the user can combine to automate work. This is
required to be done because ns -3 during its development has a adopted a coding style t o which
everybody must adhere to make contributions to the project. However, this does not seem to be
required if we just want to run simulations and gather data.
The module libraries are necessary to any user created script, hence without them the
simulat ion program will not work. They function in a similar manner to libraries in C/C++.

“#include <ns3/core -module.h>
#include <ns3/network -module.h>
#include <ns3/mobility -module.h>
#include <ns3/lte -module.h> ”

The libraries are grouped around very large modules, this is done so that the user will not have
to include many libraries in their script and deal with the problem of dependencies. It is still possible
to include individual libraries for specific modules if we do not wish to use a large include file.
Logging components are a system which main job is to print during the run of simulation
information about different modules or stats, this is done by selecting which level of the logging
component should be active. The levels of logging can be viewed in the tables below.

34
Table 3. Possible logging levels adapted from [11]
Individual Logging level Inclusive Logging Level
Severity Class Meaning Level Meaning
LOG_NONE The default, no
logging LOG_LEVEL_ERROR Only LOG_ERROR
severity class messages.
LOG_ERROR Serious error
messages only LOG_LEVEL_WARN LOG_WARN and
above.
LOG_WARN Warning
messages LOG_LEVEL_DEBUG LOG_DEBUG and
above.
LOG_DEBUG For use in
debugging LOG_LEVEL_INFO LOG_INFO and above.
LOG_INFO Informational LOG_LEVEL_FUNCTION LOG_FUNCTION and
above.
LOG_FUNCTION Function
tracing LOG_LEVEL_LOGIC LOG_LOGIC and
above.
LOG_LOGIC Control flow
tracing within
functions LOG_LEVEL_ALL All severity classes.

If one wants to see messages at single severity class is enough to use the macro settings from
the table individual logging level. On the other hand, if we wish to see messages at a desired
severity class and higher, we use the macro function from the incl usive logging level.
Prefixes are also possible to be used to help identify what, where and when is being displayed
on the screen during simulation. Prefixes can also offer a great deal of information and make easier
the job of searching for a specific log component if we use more than one. A list with prefixes and
their function is presented in the table below.

Table 4. Possible prefix values to be used , adapted from [11]
Prefix Symbol Meaning
LOG_PREFIX_FUNC Prefix the name of the calling function.
LOG_PREFIX_TIME Prefix the simulation time.
LOG_PREFIX_NODE Prefix the node id.
LOG_PREFIX_LEVEL Prefix the severity level.
LOG_PREFIX_ALL Enable all prefixes.

Usually logging components are defined at the start of the program , after the main function, to
make the work of editing and adding more logging components for different modules easier. Below
is presented how some log components are enabled and configured with the desired parameters.

“LogLevel logLevel=(LogLevel)(LOG_PREFIX_FUNC|LOG_PREFIX_TIME |LOG_LEVEL_ALL);
LogComponentEnable ("LteEnbRrc", logLevel);
LogComponentEnable ("LteUeRrc", logLevel);
LogComponentEnable ("LteEnbMac", logLevel);

The main function in ns -3 is fulfills basically the same role as in C/C++ and it declared in the
same way.

35
“int main (int argc, char *argv[])
{//the rest of the simulation program follows ”

In the case of topology helper is best to declare it before the rest of the components like nodes,
channels and net device s, because usually these entities need or are dependent on the topology
helpers.

“Ptr<LteHelper> lteHelper = CreateObject<LteHelper> (); ”

The nodes creation is straight forward, we have a class that creates nodes and next step is tell
it how many we want.

“NodeContainer enbNodes;
enbNodes.Create (numberofENBnodes);
NodeContainer ueNodes;
ueNodes.Create (numberofUEnodes); ”

To set the mobility model all we must do is instantiate an object of type MobilityHelper and
use the methods it has to install on the previous created nodes the mobility model. The below code
will place all nodes at the coordinates 0 and are stationary .

“MobilityHelper mobility;
mobility.SetMobilityModel ("ns3::ConstantPositionMobilityModel");
mobility.Install (enbNodes);
mobility.SetMobilityModel ("ns3::ConstantPositionMobilityModel");
mobility.Install (ueNodes); ”

Mobility model is a concept that define s the mobility aspects of a node. These aspects are,
whether the node is intended to be stationary or mobile and where the node is placed in the virtual
world. The mobility model of a node can be as simple as moving in one direction at given speed or
moving with a random speed in random directions.
To set the protocol stack we need first to have a net device container where to store the nodes
with the fresh installed protocol stack, we do this ope ration for both UE and eNB nodes. To install
the protocol stack, first is need ed to call the method from the previous created topology helper.

“NetDeviceContainer enbDevs;
enbDevs = lteHelper ->InstallEnbDevice (enbNodes);
NetDeviceContainer ueDevs;
ueDevs = lteHelper ->InstallUeDevice (ueNodes); ”

To attach the UEs to eNB all we need to do is use the method from the topology helper and to
specify which nodes we want to connect to the eNB.

“lteHelper ->Attach (ueDevs, enbDevs.Get (0)); ”

Now that everything is in place is time to activate the dedi cated radio bearer between UE and
eNB. This is done as before with the help of the topology helper. In this command we also tell the
bearer what QCI level we desire.

“enum EpsBearer :: Qci q = EpsBearer::GBR_CONV_VOICE;
EpsBearer bearer (q);
lteHelpe r->ActivateDataRadioBearer (ueDevs, bearer); ”

36
Now that the LTE network configuration is done , we need to tell the simulator how long we
want the simulation to last and to release all the resources used during simulation run when is
finished .

“Simulator::Stop (Seconds (0.5));
Simulator::Run ();
Simulator::Destroy ();
return 0;
}”
5.2 Handover simulation using NS -3
A first simulation scenario that has been built addresses the X2 handover as defined in section
4.8. For testing in conditions like those found in real world, several propagation models have been
selected, this propagation m odel vary from simple to more complex. This propagation models are:
Friis propagation model , two ray propagation model and three log distance propagation model.
This study was inspired from [12].
5.2.1 Study of the X2 handover procedure
The study was done using handover algorithms to enable automated handover. This handover
algorithms are implemented by ns -3 and were selected because they are very common and offer an
easy to understand process.
Both algorithms were tested using the same frequency for uplink and downlink and the same
amount RB allocated and the same frequency reuse algorithm , the same propagation model and
general layout of the network. This was done so that the results obtained will be easier to understand
and quantified. The scheme of the scenario can be seen in the figure below.

Figure 5.2 – Layout of the network for simulation

The layout of the simulation is composed out of 2 eNB s, having a distance of 1000 meters
between them, connected with a X2 interface, a remote host which acts like a substitute for internet,
the EPC of the LTE netwo rk and one UE which moves between the two cells with a speed of 20
m/s. To better understand how this scenario was implemented in ns -3 and which is the workflow
for implementing this scenario an activity flow is presented below.

37

Figure 5.3 – Handover s imulation workflow 1 of 2

Figure 5.4 – Handover simulation workflow 2 of 2

38
When the simulation starts the UE is place d initially at a distance of 100 meters from the first
eNB and starts moving towards it at a constant speed of 20 m/s , it changes signaling information
with the first eNB. As the UE approaches the second cell and starts communicating with the second
eNB the handover procedure is initiated . During the handover procedure it exchanges more
signaling information , this continues until the UE is no longer connected to the first eNB.
After the handover procedure is complete the UE continues its movement until it reaches the
distance of 100 meters from the second eNB, after it reaches this distance the simulation stops.
Another set o f tests was performed with a transmit power of 46dbm, this was done to observe
how this layout will perform at a higher power level.
In the table below are presented the simulation parameters.

Table 5. Handover s imulation parameters
Simulation Time 60s
Propagation models used Friis propagation model, Two-ray
ground -reflection, three log distance
propagation loss model
Type of traffic UDP
Number of nodes 3
Number of UEs 1
Number of eNB 2
Uplink frequency 1920 MHz
Downlink frequency 2110 MHz
Handover trigger event A2, A4, A3
Transmission power for eNB 46dbm and 20dbm
Version of IP used IPv4
Number of RB used 100
Bandwidth 20MHz
eNB antenna height 35 meters
UE antenna height 1.5 meters
eNB noise figure 5db
UE noise figure 9db
Antenna mode SISO (single antenna)
Antenna pattern omnidirectional
Serving cell threshold 30
Hysteresis (dB) 3db
Time -to-Trigger (ms) 256
Antenna gain 0
Neighbor cell offset 1
5.3 Carrier Aggregation simulation using NS -3
This chapter focuses on the simulation of carrier aggregation using the support offered by ns –
3. For simulating carrier aggregation in conditions similar to those found in the real-world
propagation models were added , namely the Friis free space propagation model and the Okumura
propagation model.
Simulation of carrier aggregation was done in two steps, at first the topology of the network is
formed out of one eNB and one stationary UE. The UE is placed first relatively close, at a distance
of 100 meters from the first eNB and after that the UE is moved further away from the eNB.

39
The second part in simulating carrier aggregation was to place the UE relatively close to eNB
and then to start moving the UE at a constant speed away fr om the eNB , this was done to illustrate
how the throughput decays in relation with the increase of distance and speed .
The traffic generator used was one which was configured do generate traffic in the downlink
direction. The amount of traffic generated was enough to sustain the theoretical speed specified for
each channel frequency and was the same across all tests . This was possible because the amount of
traffic delivered to the UE is limited by the amount of resource blocks set for the network. The
same setting s were kept for the traffic generator for all testing.
5.3.1 Study of Carrier Aggregation on stationary nodes
In the study of CA on stationary nodes was also added different transmission modes or antenna
modes and by this we refer to MIMO. The transmission modes included in the scenario are: single
antenna, transmit diversity , spatial multiplexity open loop and cl osed loop. The layout of the
network can be seen in the image below.

Figure 5.5 – Layout of the nodes

The tests performed for this topology were done with different number of RB s, namely 25 and
75. These values were chosen, because they provided the most reliable results and because the
carrier aggregation module implemented in ns -3 does not fully comply with the standards and other
number of RB are not fully supported, the cause for this is the resource allocation model from the
MAC layer , this limitation is described in more detail in [10] .
When it comes to the number of components carriers used for each level of RB, the tests were
performed with 2 and 3 component carriers at a distance of 100- and 400 -meters form eNB. To test
the effects of changing the transmission mode the UE was placed at a distance of 1500 meters form
eNB, the same propagation models and the same settings for the network were kept .
Moreover, when test ing with the UE positioned close to the edge of the cell the propagation
model Okumura Hata had more influence on throughput than Friis free space propagation model .
For testing at the edge of the cell the same setting were kept for the network. The same values for
resource blocks were used and the same transmission modes and the same component carriers . The
results obtained were different from the original. The throughput had lower values than before.
When different transmission modes were used the results varied with distance, this is because
in ns -3 different transmission modes scale different with the transmission power. The parameters
used for testing on all distances are presented in the table on the next page .

40
Table 6. Carrier aggregation s imulation parameters for stationary UE
Distance 100 and 400 meters 1500 meters
Simulation Time 10s 10s
Propagation models used Friis free space propagation model Friis free space and
Okumura Hata propagation
model
Type of traffic UDP UDP
Number of nodes 2 2
Number of UEs 1 1
Number of eNB 1 1
Uplink frequency 1920 1920
Downlink frequency 2110 2110
Bandwidth 5MHz, 15MHz 5MHz, 15MHz
Transmission power 20 dBm 20 dBm and 46 dBm
Version of IP used IPv4 IPv4
Number of RB used 25, 75 25, 75
Type of aggregation intra-band , contiguous intra-band , contiguous
Number of carriers used 2, 3 2, 3
Transmission modes used single antenna single antenna, transmit
diversity, spatial
multiplexity open loop and
closed loop
Speed of the UE Constant Position Constant Position
Antenna gain 0 0
Antenna pattern Omnidirectional Omnidirectional
eNB noise figure 5db 5db
UE noise figure 9db 9db
5.3.2 Study of C arrier aggregation on moving nodes
A second topology for this study contains a moving UE. This topology is still based on the
first topology with the added difference that the UE moves towards the edge of the cell with
constant speed. The tests were performed at three distingue values of ve locity 10 m/s, 20 m/s, 30
m/s. The testing methodology consisted out of two level of RB, namely 25 and 75, a transmission
power of 46 dBm and 20 dBm . The results were different than before, at first when the UE was
close to the eNB throughput was high and as the UE went closer to the edge of the cell it started to
decrease in value. This behavior stayed true for all level of RB and number of component carriers.
For Okumura Hata propagation model the UE was positioned at an initial distance of 1000
mete rs from eNB, to comply with requirements of the propagation model , and started moving for
a total traveled distance of 420 meters. In the case of Friis free space propagation model the re is
not any minimum required distance , it can be used regardless of the distance of the UE fr om eNB.
To be noted in the case of the free space propagation model the attenuation is relatively low
compared to Okumura Hata propagation model.
For different values of velocity, it can be observed as the rate of throughput decay increases,
for example when we increase speed from 10 m/s to 20 m/s and fr om 20 m/s to 30 m/s. A visual
representation of the layout is presented in the image on the next page .

41

Figure 5.6 – Layout of the nodes

The cause for the decay of the transfer at speeds like 10 m/s i s the increase of distance . For
higher speed like 20 m/s and 30 m/s the increase of distance coupled with the high speed cause the
throughput values to decrease faster than for 10 m/s . The throughput values depend also on the
transmission power the eNB uses . The distance traveled by UE is of 420 meters.
The parameters of the network are shown in the table below.

Table 7. Carrier aggregation s imulation parameters for moving UE
Simulation Time 50s, 25s, 16s
Propagation models used Friis propagation model, Okumura
Hata propagation model
Type of traffic UDP
Number of nodes 2
Number of UEs 1
Number of eNB 1
Uplink frequency 1920
Downlink frequency 2110
Bandwidth 5MHz, 15MHz
Traveled Distance 420 meters
Transmission power 46 dBm and 20 dBm
Antenna gain 0
Type of aggregation intra-band, contiguous
Number of RB used 25, 75
Number of carriers used 2, 3
Transmission modes used single antenna
Speed of the UE 10 m/s, 20 m/s, 30 m/s
UE noise figure 9db
eNB noise figure 5db
Antenna pattern Omnidirectional

The activity flow for studying carrier aggregation with stationary and moving UE is shown in
the figure s on the next page .

42

Figure 5.7 – CA simulation workflow 1 of 2

Figure 5.8 – CA simulation workflow 2 of 2

43
5.4 MAC schedulers simulation in NS -3
The third simulation scenario that has been built addresses the problem of MAC schedulers.
The study of MAC schedulers is done on a few MAC schedulers, namely: round robin, maximum
throughput , proportional fair, priority set scheduler and CQA scheduler.
The network layout for this study consisted out of one eNB and 3 UE s, arranged in a manner
to simulate different working modes for the schedulers. The study for all MAC schedulers was
performed with the same network characteristics and in the same network layout configurations .
The propagation models considered are Okumura Hata propagation model and Friis free space
propagation model. The propagation models were used on all schedulers and on every layout of the
network. The testing of the MAC scheduler was performed in two stages. In the first stage all UE s
had the same QoS level and the scheduler used for testing were round robin, proportional fair and
maximum throughput. In the second stage the UEs had different level of QoS set for them , more
exactly two UEs had a GBR enabled QoS level and the remaining UE had a non-GBR enabled QoS
level and the schedulers used were round robin, priority set scheduler and CQA scheduler.
The traffic set for each UE in the traffic generator was the same for all UEs and it was set such
as the sum of all traffic will be slightly higher than the maximum amount of traffic which can be
achieved for a channel frequency of 1 .4 MHz or 6 RB. This was done to make sure that the
theoretical speed for the channel frequency used is achieved, since a fine tune of the traffic
generator is not possible and also because the amount of traffic sent to the UEs is limited by number
of RBs used. These settings of the traffic generator were used for the testing on all MAC schedulers.
5.4.1 Study of scheduling mechanism on non -QoS aware schedulers
As previously stated, in this part of the study all UEs have the same QoS level, namely the
QoS level associated with conversational video or livestreaming . This part of t he study focuses on
the behavior of the MAC schedulers round robin, proportional fair and maximum throughput .
The layout of the network through the testing phase consisted out of three configurations ,
namely a configuration where all UEs are positioned in a circle like position s with the eNB in the
center , second configuration put the UEs in different positions around eNB some closer to it and
other situated farther away from it. The third and final configuration is similar to the second o ne
with the added difference that all UEs move, the UEs situated close to the edge of the cell move
towards the eNB at a constant speed and the UE placed near the eNB moved away from it at a
constant speed .
For simulating with Okumura Hata propagation model the distance between eNB and UEs was
set to be higher than 1000 meters to comply with the requirements set by the propagation model.
In the case of the Friis free space propagation model, there is not any requirement in terms of the
distance between eNB and UE.
The circle layout was meant to simulate a best -case layout where all UEs are positioned
relatively close to the eN B and with similar distances between them and similar propagation
characteristics . The layout where the UEs are placed at different distances is meant to show how
the MAC schedulers perform when one UE has better propagation characterizes than the rest of the
UEs. The layout which has mobility is meant to display the behavior of the MAC schedulers when
the UEs ar e constantly changing their position in relation to each other and eNB. The layout s of
the network are presented in the image s on the next page .

44

Figure 5.9 – UEs placed at the same distance

When the UEs are placed at the same distance from eNB is a best -case topology where all the
UEs have the same overall propagation characteristics and SINR values. This topology allows the
MAC scheduler to perform at their best without having to take into c onsideration the difference in
attention between UEs.
The second layout used for this study is more difficult for MAC schedulers as they must
provide the required throughput to UEs situated at the edge of the cell. The layout configuration is
presented in the image below.

Figure 5.10 – UEs placed at d ifferent distance s

With this layout MAC schedulers performed different from the previous circle arrangement,
the throughput towards the UEs situated further away form eNB had lower values than the UE
situated closer to the eNB . Maximum through put had the highest throughput out of the three
schedulers, followed by proportional fair and round robin.
The t hird layout used in this study was based on the second one with the added movement of
the UEs. The UE s started moving from the positions held initially, the edge UEs sta rted moving
towards the eNB at constant speed of 20 m/s and the UE situated close to th e eNB started moving
further away at constant speed of 20 m/s.

45
The throughput on each UE modified in relation with the distance and speed, as the edge UEs
come closer to the eNB the throughput increased and as the UE went further away the throughput
decrea sed for round robin and proportional fair. Maximum throughput only offered services to the
UEs that had the best connection re lative to the other UEs.
The parameters used for the LTE network and simulation are presented in the table below.

Table 8. MAC schedulers s imulation parameters
Simulation time 25s
Propagation model used Okumura Hata propagation model and Friis
free space propagation models
Number of nodes 4
Number of eNB 1
Number of UEs 3
Antenna pattern Omnidirectional
eNB noise figure 5db
UE noise figure 9db
Transmission mode Single antenna (SISO)
Traveled distance 500 meters
Channel bandwidth 1.4MHz
Number of RB used 6
Antenna gain 0
Downlink frequency 2110
Uplink Frequency 1920
Transmission power 20 dBm and 46 dBm
Speed of UEs 20 m/s
5.4.2 Study of scheduling mechanism on QoS aware schedulers
The focus of this study was to see how the QoS aware MAC scheduler s implemented in ns -3
perform and manage the available resources when there are UEs with different QoS levels. The
schedulers used were priority set and CQA, round robin was also used to o as a reference to compare
against.
The configurations for the layout of the network are the same as for the study on the previous
scheduling mechanism s. One topology were the UEs are placed around the eNB at the same
distance in a circle like formation. Second topology in which the UEs were placed at different
distance from eNB, two UEs were placed near the edge of the cell and one UE was placed near
eNB. Third topology was similar to the second topology with the added difference that the UEs
were mobile, moving at a constant speed of 20 m /s, the UE close to eNB move away and the one
furthest away moved towards eNB.
The parameters used for the network were the same for the number RB used, the frequency for
uplink and downlink, transmit power, noise figures and simulation time . The QoS levels used were
GBR_CONV_VIDEO for the UEs who had a guaranteed bit rate set for them and the other UE QoS
level set was NGBR_VOICE_VIDEO_GAMING . This two QoS levels were chosen to offer a more
accurate comparison between non -GBR and GBR enabled services. For the GBR enabled devices
the guaranteed bit rate was set to 384 Kbps .
The flow diagram which presents the workflow is presented in the images on the next page.

46

Figure 5.11 – MAC scheduler simulation workflow 1 of 2

Figure 5.12 – MAC scheduler simulation workflow 2 of 2

47
6. Experimental Results
This chapter goes through the experimental results obtained by studying X2 handover
mechanism, carrier aggregation concept and the scheduling mechanism of MAC schedulers with
the help of ns-3.
The experimental results are present in the shape of charts, graphs and screenshots obtained
during the run of the simulation. The graphs are meant to showcase aspect s related to the
functioning of the network, for example: the values of RSRP and SIN R as they change with
distance or time, values of the MCS and where the respective measurements take place in time or
in distance.
The screenshots are intended to show the reader a particular part on the functioning of a real
LTE network, but represented in ns -3, they are less focused on showing particular values and are
meant to show for example how the signaling takes place between UE and eNB during a handover,
a measurement report made by UE or any kind of signaling message used by a LTE network
emulate d in ns -3.
6.1 Handover simulation results
The f irst experimental results are for handover algorithm bases on the vents A2 and A4 the
results are divided on showing SINR, RSRP and MCS for downlink and uplink with a propagation
model added.
The propagation model used is three log distance propagation model . This propagation model
results are showcased, because they offer the most accurate representation of the conditions that
can be found in the real world from the other propagation models ch osen to be used in this study.

Figure 6.1 – Downlink values for SINR, RSRP and MCS on the A2 and A4 events

As described in the chapter focused on the implementation part, it can be seen fr om the graphs
how the values for SINR, RSRP and MCS vary with the movement of UE. At first when the UE
start moving towards the first eNB the values of RSRP and SINR increase as the UE reaches the
eNB they start decreasing until the handover is completed, a fter that they start increasing again
until the UE starts moving away from the second eNB. For MCS graph, the behavior described by

48
the graph is similar to one previously described for RSRP and SINR graph. Below are presented
the results obtained on the up link side of the connection.

Figure 6.2 – Uplink values for SINR and MCS on the A2 and A4 events

The chart above is focused on the stats present on the uplink side of the connection. The
behavior described in them is like what was described for downlink with the added difference that
the values are smaller for RSRP and SINR, this is because the transmit power of UE is smaller than
that of the eNB. The MCS chart is different than its downlink counterpart since the transmit power
of the UE is lower, the UE must change is modulation coding scheme more often than the eNB .
The second experimental results are for the handover algorithm based on A3 event, with this
algorithm the handover process takes place a little later during the simulation as it can be seen in
the charts below. The overall behavior of the network is the same as before.

Figure 6.3 – Downlink values for SINR, RSRP and MCS on the A3 based handover

On the next page are presented the results obtained in the uplink for RSRP, SINR, MCS and
some signaling information captured during the run of the simulation.

49

Figure 6.4 – Uplink values for SINR and MCS in uplink with A3 based handover

The next experimental result s presented in the images below are in the form of screenshots and
are meant to show the signaling procedure for a handover procedure as presented in section 4.8 in
the chapter dedicated to theoretical fundamentals . The screenshots were taken with the help of
logging components tool during simulation run.

Figure 6.5 – Original eNB handover preparations

Figure 6.6 – Handover request and handover acknowledge exchange

In the images above, it can be seen how the source eNB prepares for handover at a timestamp
of 30.612928570s, next in the timeline of the simulation the target eNB almost instantly receive
the handover request at a timestamp of 30.61292880s and and sends back the handover request
acknowledge to the source eNB.

50
Once the source eNB receive the handover ack form the target eNB the handover procedure
officially starts and more signaling information is excha nged between the two entities. This
exchange of messages is shown in the image below.

Figure 6.7 – Handover process progress

To conclude the handover procedure, it is needed for the path switch signal to be sent to the
MME and the release context signal to be sent from the target eNB to the source eNB and so the
tunnel which connected the UE to the S -GW moves form the original eNB to the new eNB after
the handover is completed . Once all of this is done the RRC reconfiguration is done. All of this is
shown in the image below.

Figure 6.8 – Handover procedure finalization
6.2 Carrier Aggregation simulation results
This chapter covers how throughput on a UE varies when carrier aggregation is enabled
compared when carrier aggregation in not enabled on the same UE. How the throughput varies
with the change of transmission mode and when the UE is mobile . A better explanation of the
testing methodology and overall layout of the network is presented in the chapter focused on
implementation.
6.2.1 Carrier Aggregation results on stationary UE
First experimental results are focused on showing the results obtained at a distance of 100
meters. The results in the figure s on the next page are obtained from running the simulation with
Friis free space propagation model enabled . At 100 meters the propagation model do es not have
much of an influence.

51

Figure 6.9 – Throughput values obtained at 100 meters using a 5MHz on each component carrier

In the above figure one can see a clear difference in terms of throughput between a UE with
no carrier aggregation and one with carrier aggregation enabled. There is also is a difference
between a UE with carrier aggregation on 2 component carriers versus one with 3 component
carriers on 5 MHz or 25 resource blocks . In this case the MCS index for the UE has the values of
28, as can be seen in the image below.

Figure 6.10 – Output file MCS index in the downlink direction

The MCS index obtained is due to the great SINR resulted from the close proximity of the UE
to the eNB. In the image below are presented the results obtained in the same conditions as before,
but the channel frequency was changed to 15 MHz or 75 resource blocks.

Figure 6.11 – Throughput values obtained at 100 meters using 15MHz on each component carrier

52
One can deduct from the previous two images presented on the previous page that the
throughput values increase percentage from the change of number of component carriers is higher
for the case in which was used 25 resource blocks than one with 75 resource blocks. At a distance
of 400 meters fr om eNB the throughput values are similar to those obtained for a distance of 100
meters. In the image below are presented the effects that the change of transmission mode has on
the throughput on CA enabled UE.

Figure 6.12 – Different transmission modes throughput values at 1500 meters

The figure above shows how the throughput values increase when the transmission mode is
changed from single antenna to transmit diversity on the same UE at a distance of 1500 meters
form eNB on free space propagation model on 5Mhz or 25 resource blocks on each component
carrier . The sam e behavior stays true for 15MHz or 75 resource blocks . Below are presented the
throughput values for spatial multiplexing transmission mode.

Figure 6.13 – Spatial multiplexing throughput values at 1500 meters

In the case of spatial multiplexing transmission mode, the values for throughput are higher
than those obtained for single antenna and transmit diversity in the same conditions and channel

53
bandwidth of 5MHz . One can observe in the image presented that the values obtained for spatial
multiplexing are very similar for open loop and closed loop .
6.2.2 Carrier Aggregation results on moving UE
Next it will be presented how the carrier aggregation performs when the UE is moving at two
different values of velocity. A detailed explanation of this study is done in the chapter on
implementation. The graphs presented in this section are focused on the showcasing how the values
of throughput vary with the mobility of the UE at speed of 10 m/ s and 30 m/s. The propagation
model used is Okumura Hata.
The graph below shows the values obtained for a UE which is positioned initially at a distance
of 1000 meters form eNB and then start to move away at a constant speed for 420 meters.

Figure 6.14 – 10 m/s throughput values

On the figure below are presented the results obtained at speed of 30 m/s.

Figure 6.15 – 30 m/s throughput values

In the images presented above one can observe that the same process stays true for both values
of velocity regardless of the number of resource blocks and component carriers used , the process
which causes the throughput values to change in relation with distance is link adaptation.

54
6.3 MAC scheduling mechanism results
This chapter covers how different MAC scheduler s scheduling mechanism s behave in a
simulation environment in ns -3. The results of the MAC scheduling mechanism are divided into
two categories. The first category is fo r MAC schedulers which do not consider the Qo S parameters
and the other is for MAC schedulers which take into consideration the QoS.
The simulated environment in which the result s were obtained is made from three layouts ,
namely first layout were all U Es are positioned at the same distance from eNB, second layout in
which the U Es were positioned at different distances and third layout w here the U Es could move
at a constant speed. A more detailed explanation is available in the chapter dedicated to
implementation.
6.3.1 MAC scheduling results on non-QoS aware schedulers
The results in this section were obtained with the Okumura Hata propagation model , the UEs
were placed at distances gr eater than 1000 meters form eNB, more precise at a distance of 1200
meters for same distance configuration and for the other two topologies the distance was increased
to 1500 meters . This propagation model was used for all layout s used in the s imulation. The results
are present in the form of a chart s and are m eant to show how the MAC scheduler s used in the
simulation compare to each other.
First results are for the layout where all the U Es are placed around the eNB at the same
distance , the results are presented in the image below. The results showcase how each MAC
scheduler manages each UE.

Figure 6.16 – Same distance throughput values for each scheduler

Even though all the U Es were place d at the same distance around the eNB, each scheduler
manages the resources of the network in a different manner. Round robin tries to share the available
resources equally among U Es, proportional fair values are not much different from round robin this
is because the overall attenuation on U Es is similar . Maximum throughput is not shown on this
graph because when the UEs hav e the same achievable rate or present the same propagation
characteristics , current implementation of maximum throughput in ns -3 will always selects the first
UE created in scrip t.

55
The second set of results are for the layout in which the U Es are positioned at different
distances fr om the eNB and are stationary. The results are present in the image on the next page .

Figure 6.17 – Different distance throughput values for each scheduler

The values for each MAC scheduler have changed from before. Round robin has the same
behavior as before, proportional fair offers slightly higher overall throughput than round robin
while servicing all U Es. Maximum throughput tries to maximize the maximum throughput of the
eNB, it does this by offering the highest possible throughput on the UE which can achieve the
maximum pos sible data rate and in this case is UE -2, being the closest to the eNB.
Next results are for the case in which the U Es move, the U Es start moving from different
positions . The chart s presented in the image below are for maximum throughput scheduler.

Figure 6.18 – Maximum throughput results

56
Since the U Es are positioned at different positions around eNB the maximum throughput
scheduler will maximize the throughput on the U Es which are closer to eNB and ignore those which
are further away. As the U Es move they’re positions change relatively to the eNB, this change of
positions determines which U Es are serviced by the network according to maximum throughput
scheduler. Next image showcases the behavior of round robin and proportional fair scheduler .

Figure 6.19 – Round robin and proportional fair results

The behavior of the schedulers presented above show how both round robin and proportional
fair offer similar values on the U Es, although proportional fair o ffers better and more stable values
overall .
6.3.2 MAC schedul ing results on QoS aware schedulers
This section covers the results for the MAC schedulers that take into consideration the QoS
parameter in their scheduling mechanism , UE-1 and UE -2 are set to have a GBR enabled QoS level
and UE -3 is set to have a non – GBR enabled QoS level .
The results presented are for round robin, priority set and CQA scheduler. Round robin is used
as reference to compare with the other two QoS aware MAC schedulers. The image on the next
page presents the results for the first layout where the U Es are placed at the same distance around
eNB.

57

Figure 6.20 – Same distance UE position results

In the case of round robin and priority set schedulers b y placing the U Es at the same distance
it eliminates most of the factors that influence the throughput which in turn allows the schedulers
to offer the maximum possible throughput to each UE regardless of their QoS setting. CQA
behavior resembles that of the maximum throughput scheduler but focused on UEs with GBR set
for them . CQA scheduler first share s the resources among the UEs who have guaranteed bit rate
set for them and will try to maximize the maximum transfer rate for them beyond the value set for
guaranteed bit rate. The figure below showcases how the sched uler manages the resources when
the U Es are placed at different distances.

Figure 6.21 – Different distance UE position results

In the chart above it can be clearly be seen a difference between the schedulers. Round robin
shares all the available resources between the UEs . Priority set , since is a scheduler which takes
into consideration the guaranteed bi t rate parameter will try t o keep the values of the throughput

58
around that specific values for UE -1 and UE -2, for UE -3 the value of the throughput is much lower
because it does not have GBR value set for it. CQA behavior is the same as before when the UEs
were placed at the same dis tance around the eNB. Although is possible for CQA scheduler to
service the third UE if any resources remain unused from the other two UEs.
The results obtained after the mobility was added to the UEs are show n in the image below .

Figure 6.22 – Round robin , priority set and CQA mobility results

The graph above presents the behavior of each scheduler when mobility of the UEs is added
into the mix. The behavior of round robin is predictable with increases and decreases of the
throughput values in relation with distance for all UEs , in this case UE-1 and UE -3 share the same
throughput growth overall . Meanwhile priority set tries to keep the throughput for UE-1 and UE -2
around the value set for guaranteed bit rate and for UE -3 resources are allocated but is not
considered a priority until the propagation conditions allows for higher throughput.
CQA only services the UEs who have a guaranteed bit rate set for them, its behavior is similar
to the one characteristic to maximum throughput schedu ler but focused only on UEs with GBR
parameter set for them. It will try to maximize the throughput beyond the values set for GBR even
if it means that non – GBR UEs will not have any resources allocated to them. This behavior stays
true until resources ca n be allocated to the non – GBR enabled UEs.

59
7. Conclusion
As previously stated, the purpose of this work was the study of the LTE network with the help
of the ns-3 simulation platform. The study of the LTE network performed in this work is not an
exhaustive one, this study in essence focuses on some key aspects of an LTE and LTE-A network.
The aspects covered are in the field of data transfer and mobility suppor t on the radio interface
and are mean t showcase the working of this processes in a virtual environment in specific
conditions. Given these points the process studie d are those of the handover mechanism, carrier
aggregation and MAC scheduling mechanism.
The study of the handover mechanism was focused in showing how typical handover event
will proceed in a simple configuration composed out of two eNB and one moving U E in virtual
environment. The simulation process was a successful one , it was possible to generate and trigger
signaling information specific to handover event and to see specific measurement measurements
performed by the network and how these measurements i nfluence the handover process.
It was also possible to see signaling information or messages for different components,
protocols and action s performed by the network by using logging compo nents .
For the study of carrier aggregation, it was possible to perform many simulations with different
number of component carriers , channel bandwidth s and different transmission modes. Although i t
was possible to study varied amount of configuration, it was necessary to limit the study to only
some of th em and to a specific environment . This limitation was caused by the fact tha t the module
used for simulating carrier aggregation does not completely adhere to theory and it is not certified
by any official authority.
Even with these limitations imposed, this work manages to display t he basic functionality of
carrier aggregation as a concept and to demonstrate the usefulness that this concept brings to an
LTE network.
The research of the scheduling mechanism characteristic to MAC schedulers focused on
showcasing how basic and more advanced MAC scheduler function in an LTE ne twork. The study
managed to present how these schedulers behave and to compare them with each other to better
understand the differences that are stated in theory for them. While the study focuse s on simple
network topolog ies, it still succeed s in showing the working of the scheduling mechanism.
In conclusion , this work show s that ns -3 can be used as a tool for simulating and studying a
complex network as LTE . NS-3 as a tool is not limited to only simulating mobile networks, it can
be used in the study of multitude of other fields. Overall ns -3 is mainly suited for research and
academic projects, since it is developed as an open source software.
The open source nature of ns -3 brings with it some limitations, that depending on the field of
research or purpose of the project can make the software inappropriate for use. This limitation
comes as a consequence of the fact that ns-3 is not backed by any big corporation or by any official
authority.
As future work for this thesis can be considered the continuous research of other aspects of the
LTE network and in further study of the already touched subjects in both terms of topology
complexity and interworking with other components of the LTE network. By doing this is possible
to use ns -3 to its full capacity.