Modeling and Simulation of Scheduling Algorithms in LTE Networks

Academic Year 2011/2012

ELECTRICAL AND COMPUTER ENGINEERING

THE INSTITUTE OF TELECOMMUNICATIONS

FACULTY OF ELECTRONICS AND INFORMATION TECHNOLOGY

WARSAW UNIVERSITY OF TECHNOLOGY

Bachelor of Science Thesis

Modeling and Simulation of Scheduling Algorithms

in LTE Networks

Dinesh Mannani

Supervisor:

Dr. Mirosaw Somiski, Associate Professor

Consultant:

Dr. Sawomir Pietrzyk, IS-Wireless

...........................................................

Evaluation

...........................................................

Signature of the Head

of Examination Committee

Warsaw, January 2012

2

Modelling and Simulation of Scheduling Algorithms in

LTE Networks

Abstract:

This thesis is based on the study of scheduling algorithms in LTE (Long Term Evolution).

LTE is an evolution of the UMTS (Universal Mobile Telecommunications System)

standardised by the 3GPP (3rd

Generation Partnership Project) in its Rel. 8 for the

development of wireless broadband networks with very high data rates. It enables mobile

devices such as smartphones, laptops, tablets to access internet at a very high speed data

along with lots of multimedia services. The future of LTE lies in being implemented in

various electronic devices to exchange data wirelessly at very high speeds.

Technically, the Long Term Evolution provides a high data rate and can operate in different

bandwidths ranging from 1.4MHz up to 20MHz. In terms of features the latest Release of

LTE (Rel. 10 LTE-Advanced) aims to deliver enhanced peak data rates to support advanced services and applications (100 Mbit/s for high and 1 Gbit/s for low mobility [25], low latency

(10ms round-trip delay), improves system capacity and coverage, supports multi-antenna and

reduces operating costs [1] by introducing concepts like SON and allowing seamless

integration with existing mobile network systems.

Scheduling is basically the process of making decisions by a scheduler regarding the

distribution of resources (time and frequency) in a telecommunications system among its

users. The Max SNIR, the Proportional Fair and the Round Robin scheduling algorithms

have been considered and discussed in this dissertation. The analysis of these scheduling

algorithms has been done through simulations executed on a MATLAB-based system level

simulator from IS-Wireless called LTE MAC Lab (aka Matlab version of 4G System Lab)

with their verification in a LTE network test environment (deployed in the Institute of

Telecommunications within the Smart City of TPSA in the Warsaw University of Technology).

I have examined the impact on the throughput and the fairness results of each scheduling

algorithm.

This thesis is mainly to understand the scheduling algorithms for LTE by means of modelling

and simulation of the process and in the end verify the results by conducting tests in a LTE

test environment. Furthermore, work out a method to examine LTE scheduling performance

evaluation for teaching purposes.

3

Modelowanie i symulacja dziaania procedur rezerwacji zasobw w sieciach typu LTE

Streszczenie:

W pracy podjto studia dotyczce nowoczesnych sieci telekomunikacyjnych zgodnych ze standardem Long Term Evolution ( LTE), opracowanym i rozwijanym przez Konsorcjum 3rd

Generation Partnership Project (3GPP), ktre pozwalaj ju obecnie na osiganie w sieciach telefonii komrkowej duych szybkoci transferu danych (do 100 Mb/s w kierunku do abonenta i do 50 Mb/s w kierunku zwrotnym) z maymi opnieniami. W szczeglnoci skupiono si na zagadnieniach zwizanych z analiz procesu rezerwacji zasobw transmisyjnych w tych sieciach.

Do analizy porwnawczej wybrano trzy, rekomendowane dla tych sieci, algorytmy: The Max

SNIR Scheduling Algorithm, The Proportional Fair Scheduling Algorithm i The Round Robin

Scheduling Algorithm. Analizy przeprowadzono z wykorzystaniem profesjonalnego narzdzia programistycznego Symulatora LTE MAC Lab (aka Matlab version of 4G System Lab) firmy IS-Wireless oraz niedawno otwartego na Wydziale Elektroniki i Technik

Informacyjnych (WEiTI) Politechniki Warszawskiej Laboratorium LTE, przygotowanego we

wsppracy z firm HUAWEI Polska Sp. z o.o., Telekomunikacj Polsk S.A. i Orange Labs Poland.

Uzyskane w pracy wyniki zostan wykorzystanie w zajciach dydaktycznych na studiach 1- i 2-stopnia WEiTI prowadzonych z wykorzystaniem ww. Laboratorium LTE oraz badaniach

i testach wykonywanych przez studentw w rodowisku sieciowym Miasteczka Testowego Telekomunikacji Polskiej w Politechnice Warszawskiej.

4

CURRICULUM VITAE

Personal Details: Name: Dinesh Mannani

Date of Birth: 28-02-1990

Nationality: Indian

Work Experience: 1. Intern Business Development Team at ISWireless, Warsaw, Poland

01-07-2011 to 30-09-2011 2. English Teacher (Native)

Since 03-2011

Career Highlights: 1. School Prefect

Head of the Student Union

Represented school at various public events

worked in a team environment to represent students 2. President Computer Club at School

Responsible for handling web-designing projects

Responsible for time-management of other participants 3. President English Literary Society at School

Represented the school at various debate competitions 4. Experience in Hospitality Industry

5. Experience in website and graphics designing

Education: Completed schooling from Birla Vidya Mandir, Nainital, India

o 04-2004 to 03-2008

BSc. in Electrical and Computer Engineering (specialization Telecommunications) at Warsaw University of Technology (Politechnika Warszawska).

Thesis: Scheduling Algorithms in LTE networks o 10-2008 to 02-2012

Skills: Knowledge of LTE/WiMax network technologies.

Knowledge of various Routing and internet protocols.

Knowledge about Business Development and CRM

Fluent in English, Hindi.

Working knowledge of Polish.

Working knowledge of Visual Basic, HTML, JavaScript, flash, SQL.

Designing magazines, graphics etc.

.. Signature of the student

5

Acknowledgement

This Bachelors thesis is the final step in obtaining my Bachelors Degree in Electrical and Computer Engineering with specialisation in Telecommunications at the Warsaw University

of Technology.

The thesis was conducted under the supervision of Dr. Mirosaw Somiski, Associate Professor in the Telecommunications Department of the Faculty Electronics and Information

Technology at the Warsaw University of Technology (Politechnika Warszawska). I have

worked on my Bachelors thesis from June, 2011 to January 2012. Here I would like to express my sincere gratitude to all those who have provided me with encouragement and

guidance during this thesis.

First of all I am particularly indebted to Dr. Mirosaw Somiski, my supervisor. He has been a great support since the beginning of the thesis and showed trust in me when I first

approached him with the aim of finishing my thesis within one working semester. In some

circumstances where I had some unexpected problems during my project he was there to find

a solution and provide useful guidance. Further, I want to express my gratitude to

Dr. Sawomir Kukliski, who was always ready to share his knowledge and experience in the field of LTE.

Secondly, I would like to thank Dr. Sawomir Pietrzyk CEO of IS-Wireless and his team, for lending me invaluable knowledge support along with granting a trial license for their tool

LTE MAC Lab. Their help and suggestions have proved as important as the license itself. I

especially would like to thank Mr. Marcin Dryjaski a specialist with IS-Wireless, who has been constantly providing me with concrete suggestions on working with the thesis along

with answering all questions that I had.

6

Table of Contents 1. Introduction .......................................................................................................................... 11

1.1 Background .................................................................................................................... 11

1.2 Motivation and goals of the thesis ................................................................................. 12

1.2.1 Motivation ............................................................................................................... 12

1.2.2 Thesis goals ............................................................................................................. 13

1.3 Thesis Scope .................................................................................................................. 13

2. An Overview of LTE ........................................................................................................... 14

2.1 LTE requirements .......................................................................................................... 14

2.2 Multiple Access Techniques .......................................................................................... 15

2.2.1 Downlink - Orthogonal Frequency Division Multiple Access (OFDMA) ............. 15

2.2.2 Uplink - Single Carrier - Frequency Division Multiple Access (SC-FDMA) ........ 16

2.3 LTE Frame Structure ..................................................................................................... 17

2.4 LTE Downlink Physical Channels ................................................................................. 18

2.5 LTE Uplink Physical Channels ...................................................................................... 20

2.6 Multiple Input Multiple Output ..................................................................................... 21

3. Selected Issues of Scheduling .............................................................................................. 23

3.1 Selected Scheduling Algorithms .................................................................................... 24

3.1.1 Round Robin Scheduling ........................................................................................ 24

3.1.2 Max SNIR Scheduling ............................................................................................ 25

3.1.3 Proportional Fair Scheduling .................................................................................. 26

4. Simulations and Testing ....................................................................................................... 27

4.1 LTE MAC Lab System Level Simulator: An overview ................................................ 27

4.1.1 Simulation Scenarios .............................................................................................. 27

4.1.2 Simulation Results and Analysis ............................................................................ 28

4.2 LTE network test environment ...................................................................................... 50

4.2.1 Testing Scenarios .................................................................................................... 50

4.2.2 Testing Results and Analysis .................................................................................. 50

5. A Student Lab Experiment................................................................................................... 57

5.1 Simulation Tools ............................................................................................................ 57

5.2 Investigation of scheduling algorithms with LTE MAC Lab Matlab tool..................... 57

5.3 Summary of abilities to be gained during the experiment ............................................. 58

6. Conclusions and future work ............................................................................................... 59

6.1 Conclusion ..................................................................................................................... 59

6.2 Future work .................................................................................................................... 60

7. References ............................................................................................................................ 61

7.1 CD contents .................................................................................................................... 62

7

Figures List

Fig. 1 : OFDM and OFDMA [28] ............................................................................................ 16 Fig. 2 : OFDM and SC-FDMA [28] ........................................................................................ 16 Fig. 3 : LTE frame structure [18] ............................................................................................. 17 Fig. 4 : Frame Type 2 [27] ....................................................................................................... 18

Fig. 5 : LTE Downlink channels [18] ...................................................................................... 19 Fig. 6 : LTE Uplink Channels [18] .......................................................................................... 20 Fig. 7 : Single user MIMO transmission principle [8] ............................................................. 22 Fig. 8 : Multi-user MIMO transmission principle [8] .............................................................. 22 Fig. 9 : Layer 2 functionalities for dynamic packet scheduling, link adaptation, and HARQ

Management [8] ....................................................................................................................... 23 Fig. 10 : Flow Chart for Round Robin Algorithm ................................................................... 24

Fig. 11 : Flow chart for Max SNIR algorithm ......................................................................... 25

Fig. 12 : Flow chart for Proportional Fair Algorithm .............................................................. 26 Fig. 13 : A tree diagram for all the scenarios under consideration for simulations ................. 28 Fig. 14 : PRB allocation based on SNIR values for single user downlink Case 1................... 29 Fig. 15 : Resource Allocation for a single user in downlink Case 1 ........................................ 30

Fig. 16 : Throughput Results for single user in downlink Case 1 ............................................ 30 Fig. 17 : PRB allocation based on SNIR values for 3 users .................................................... 31

Fig. 18 : Resource allocation by RR algorithm for 3 users in downlink Case 2 ...................... 31 Fig. 19 : Resource allocation by Max SNIR algorithm for 3 users in downlink Case 2 .......... 32 Fig. 20 : Resource allocation by Max SNIR algorithm for 3 users in downlink Case 2 ......... 32

Fig. 21 : Comparison of PRB allocation in all three algorithms over time Case 2 .................. 33 Fig. 22 : Comparison of throughput obtained from all three algorithms Case 2 ..................... 33

Fig. 23 : Resource allocation by RR algorithm for 3 users in downlink Case 3 ...................... 34

Fig. 24 : Resource allocation by Max SNIR algorithm for 3 users in downlink Case 3 .......... 35

Fig. 25 : Resource allocation by PF algorithm for 3 users in downlink Case 3....................... 35 Fig. 26 : Comparison of PRB allocation in all three algorithms over time Case 3 .................. 36

Fig. 27 : Comparison of throughput obtained from all three algorithms Case 3 ..................... 36 Fig. 28 : Resource allocation by RR algorithm for 3 users in downlink Case 4 ...................... 37 Fig. 29 : Resource allocation by Max SNIR algorithm for 3 users in downlink Case 4 .......... 38

Fig. 30 : Resource allocation by PF algorithm for 3 users in downlink Case 4....................... 38 Fig. 31 : Comparison of PRB allocation in all three algorithms over time downlink Case 4.. 39 Fig. 32 : Comparison of throughput obtained from all three algorithms downlink Case 4 ..... 39

Fig. 33 : PRB allocation based on SNIR values for single user in uplink Case 1 ................... 40 Fig. 34 : Resource Allocation for a single user in uplink Case 1 ............................................. 41 Fig. 35 : Throughput results of single user in uplink Case 1 ................................................... 41 Fig. 36 : PRB allocation based on SNIR values for 3 users .................................................... 41 Fig. 37 : Resource allocation by RR algorithm for 3 users in uplink Case 2 ........................... 42

Fig. 38 : Resource allocation by Max SNIR algorithm for 3 users in uplink Case 2 .............. 42 Fig. 39 : Resource allocation by PF algorithm for 3 users in uplink Case 2 ........................... 43

Fig. 40 : Comparison of PRB allocation in all three algorithms over time uplink Case 2 ...... 43 Fig. 41 : Comparison of throughput obtained from all three algorithms uplink Case 4 .......... 44

Fig. 42 : Resource allocation by RR algorithm for 3 users in uplink Case 3 ........................... 45 Fig. 43 : Resource allocation by Max SNIR algorithm for 3 users in uplink Case 3 .............. 45 Fig. 44 : Resource allocation by PF algorithm for 3 users in uplink Case 3 ........................... 46 Fig. 45 : Comparison of PRB allocation in all three algorithms over time uplink Case 3 ...... 46 Fig. 46 : Comparison of throughput obtained from all three algorithms uplink Case 3 .......... 47

8

Fig. 47 : Resource allocation by RR algorithm for 3 users in uplink Case 4 ........................... 48

Fig. 48 : Resource allocation by Max SNIR algorithm for 3 users in uplink Case 4 .............. 48 Fig. 49 : Resource allocation by PF algorithm for 3 users in uplink Case 4 ........................... 48 Fig. 50 : Comparison of PRB allocation in all three algorithms over time uplink Case 4 ...... 49 Fig. 51 : Comparison of throughput obtained from all three algorithms uplink Case 4 .......... 49

Fig. 52 : Throughput results from Download within 5m of eNodeB: 100 MB file ................ 51 Fig. 53 : Throughput results from Download within 5m of eNodeB: 200 MB file ................ 51 Fig. 54 : Throughput results from Download within 5m of eNodeB: 500 MB file ................ 52 Fig. 55 : Throughput results from Download within 5m of eNodeB: 1 GB file ..................... 52 Fig. 56 : Throughput results from HTTP Download with user 3 at cell edge ......................... 53

Fig. 57 : Throughput results from HTTP Download with user 1 & 3 at cell edge .................. 53 Fig. 58 : Throughput results from HTTP Download with all 3 users at cell edge ................... 54 Fig. 59 : Throughput results from FTP Download within 5m of eNodeB : 500 MB file ........ 54 Fig. 60 : Throughput results from FTP Download with user 3 at cell edge ............................ 55

Fig. 61 : Throughput results from FTP Download with user 1 & 3 at cell edge ..................... 55 Fig. 62 : Throughput results from FTP Download with all 3 users at cell edge ...................... 56

Tables List

Table 2.1 : Bandwidth and Resource blocks specifications [1] ............................................... 18

Table 4.1 summary of simulation parameters used for all the testing scenarios ..................... 28 Table 4.2 : LTE Test Environment Test 1 ............................................................................... 51

Table 4.3 : LTE Test Environment Test 2 ............................................................................... 51 Table 4.4 : LTE Test Environment Test 3 ............................................................................... 52 Table 4.5 : LTE Test Environment Test 4 ............................................................................... 52

Table 4.6 : LTE Test Environment Test 5 ............................................................................... 53

Table 4.7 : LTE Test Environment Test 6 ............................................................................... 53 Table 4.8 : LTE Test Environment Test 7 ............................................................................... 54 Table 4.9 : LTE Test Environment Test 8 ............................................................................... 54

Table 4.10 : LTE Test Environment Test 9 ............................................................................. 55 Table 4.11 : LTE Test Environment Test 10 ........................................................................... 55

Table 4.12 : LTE Test Environment Test 11 ........................................................................... 56

9

Abbreviations

3GPP 3rd Generation Partnership Project LTE Long Term Evolution MMOG Multimedia Online Gaming HSPA High Speed Packet Access 3G Third Generation of Cellular Wireless Standards GSM Global System for Mobile Communication UMTS Universal Mobile Telecommunications System UTRA UMTS terrestrial radio access E-UTRA Evolved UMTS terrestrial radio access UTRAN UMTS Terrestrial Radio Access Network E-UTRAN Evolved UMTS Terrestrial Radio Access Network MIMO Multiple Input Multiple Output FDD Frequency Division Duplex TDD Time Division Duplex OFDM Orthogonal Frequency Division Multiplexing OFDMA Orthogonal Frequency Division Multiple Access SC-FDMA Single Carrier Frequency Division Multiple Access FDMA Frequency Division Multiple Access PAPR Peak to Average Power Ratio BS Base Station eNodeB Base Station MS Mobile Station UE User Equipment RB Resource Block RE Resource Element SNIR Signal to Noise-Interference Ratio RR Round Robin PF Proportional Fair CQI Channel Quality Indicator TPSA Telekomunikacja Polska S.A. DL Downlink UL Uplink HSDPA High Speed Downlink Packet Access C.D.F Cumulative Distribution Function EUL Enhanced Uplink SC Single Carrier SISO Single Input Single Output MME Mobility Management Entity SGW Serving Gateway PGW PDN Gateway CP Cyclic Prefix DwPTS Downlink Pilot Time Slot GP Guard Period UpPTS Uplink Pilot Time Slot PBCH Physical Broadcast Channel PCFICH Physical Control Format Indicator Channel PDCCH Physical Downlink Control Channel

10

PHICH Physical Hybrid ARQ Indicator Channel HARQ Hybrid Automatic Retransmission Request RS Reference Signal CIR channel impulse response PRN pseudorandom number P-SS and S-SS Primary and Secondary Synchronization Signal DC Dedicated Control ACK Acknowledge NACK Not Acknowledge PDSCH Physical Downlink Shared Channel PMCH Physical Multicast Channel PUCCH Physical Uplink Control Channel PUSCH Physical Uplink Shared Channel PRACH Physical Random Access Channel WCDMA Wideband Code Division Multiple Access PHY Physical layer MAC Medium Access Control RLC Radio Link Control RRC Radio Resource Control IEEE Institute of Electrical and Electronics Engineers 4G Fourth Generation of Cellular Wireless Standards SNR Signal to Noise Ratio PS Packet Scheduler TTI Transmission Time Interval MCS Modulation and Coding Scheme QoS Quality of Service AMC Adaptive modulation and coding PRBs Physical Resource Blocks RRM Radio Resource Management CCI co-channel interference VoIP Voice over Internet Protocol QPSK Quadrature Phase-Shift Keying QAM Quadrature Amplitude Modulation

11

1. Introduction This chapter is dedicated to the introduction to the concept of 3

rd Generation Partnership

Project (3GPP) Long Term Evolution (LTE) and technological features associated with it.

The first section, 1.1, of this chapter will discuss the background information on the subject

of LTE and scheduling. The motivation and goals for the thesis have been discussed in the

sections 1.2 and 1.3 respectively with the last section 1.3 presenting the outline of the thesis.

1.1 Background

Over the recent years we have seen mobile broadband become a reality as more and more

internet users are getting accustomed to having broadband access wherever they go, and not

just at home or in the office. Multimedia applications such as Multimedia Online Gaming

(MMOG), mobile TV, Web 2.0, streaming contents through the Internet have gathered more

attention by the internet generation and have motivated the 3GPP to work on the LTE which

is a successor to High Speed Packet Access (HSPA) currently being used in the 3rd

Generation of Cellular Wireless Standards (3G) networks. LTE is an answer to deliver better

applications and services to mobile users which consume a lot of bandwidth.

The 3GPP is the organisation which stipulates and standardises the specifications for LTE

along with Global System for Mobile Communication (GSM) and 3G Universal Mobile

Telecommunications System (UMTS) terrestrial radio access (UTRA) systems. It started

work on the evolution of 3G mobile system in November 2004, and the project came to be

known as LTE. The main focus of this initiative to introduce LTE was on enhancing the

UTRA and optimizing 3GPPs radio access architecture. A lot of research has been carried out since 2004 and proposals have been presented on the evolution of the UTRAN. The

specifications related to LTE are formally known as the evolved UMTS terrestrial radio

access (E-UTRA) and evolved UMTS terrestrial radio access network (E-UTRAN), but are in

general referred as project LTE.

The end of year 2008 saw the Release 8 of the 3GPP, which cites the stable specifications for

LTE, being frozen. The initial deployment of LTE began in 2010 with many operators

adopting it gradually. According to Release 8 specs, LTE supports peak rates of 300Mb/s

which could be achieved with the help of Multiple Input Multiple Output (MIMO) and a

radio-network delay of less than 5ms. In addition to that it operates on both Frequency

Division Duplexing (FDD) and Time Division Duplexing (TDD) and can be deployed in

different bandwidths depending on the availability of spectrum. In TDD configuration the

uplink and downlink operate in same frequency band whereas with FDD configuration the

uplink and downlink operate in different frequency bands.

Orthogonal Frequency Division Multiplexing (OFDM) has been adopted as the downlink

transmission scheme for the 3GPP LTE [11]. The transmission which occurs from the base

station to the User Equipment is referred to as downlink whereas vice-versa uplink. OFDM

divides the transmitted high bit-stream signal into different sub-streams and sends these over

many different/parallel sub-channels. For uplink transmission scheme the 3GPP selected SC-

FDMA (Single Carrier Frequency Division Multiple Access). An uplink is a transmission from the mobile station to the base station. SC-FDMA is a modified form of Orthogonal

Frequency Division Multiple Access (OFDMA) and has similar throughput performance and

12

essentially the same overall complexity as OFDMA. Like OFDM, SC-FDMA also consists of

sub-streams but it transmits on sub-channels in sequence not in parallel which is the case in

OFDM, which prevents power fluctuations in SC-FDMA signals i.e. low Peak to Average

Power Ratio (PAPR). A base station (BS) is called an Evolved NodeB (eNodeB) in the Long

Term Evolution and a mobile station (MS) is called a User Equipment (UE) in the Long Term

Evolution.

The data transmission in LTE is organized as physical resources which are represented by a

time-frequency resource grid consisting of Resource Blocks (RB). Resource blocks consist of

a no. of Resource Elements (RE). One of the major functionalities that have been assigned to

the BS is scheduling which is carried out by scheduler. The scheduler is responsible for

assigning the time and frequency resources to the different UE under the BS coverage. It does

that by allotting the RBs which are the smallest elements that can be assigned by a scheduler.

In the thesis we will be discussing the major scheduling algorithms that are used by the

schedulers, they are, Max Signal to Noise-Interference Ratio (SNIR) Scheduling, Round

Robin (RR) Scheduling and Proportional Fair (PF) Scheduling. In brief, the Max SNIR

scheduling assigns the resource blocks to the user with the highest Channel Quality Indicator

(CQI-received as a feedback from the UE by the BS) on that RB. In Round Robin scheduling

the UEs are assigned the resource blocks in turn (one after another) without taking the CQI

into account, allocating resources to the users equally. In Proportional Fair Scheduling the

UEs are assigned the resource blocks on the basis of the best relative channel quality i.e. a

combination of CQI & level of fairness desired.

The Max SNIR Scheduling, RR Scheduling and PF Scheduling have been simulated in a

MATLAB-based System Level Simulator (LTE MAC Lab) from IS-Wireless. The

performance of these scheduling algorithms in terms of throughput is analysed. We have

considered various scenarios for proper analysis in the thesis. Furthermore, the algorithms

have been analysed with their implementation in an LTE network test environment (deployed

in the Institute of Telecommunications within the Smart City of TPSA in the Warsaw

University of Technology).

1.2 Motivation and goals of the thesis

1.2.1 Motivation

The rise of the wireless industry in the past years along with the innovation of technologies

bringing large amount of multimedia services to the mobile devices led me to work in a field

where I could be a part of this wireless revolution. With LTE the future of mobile broadband

becomes brighter and clearer. According to various statistics, LTE would be the leading

technology to serve mobile broadband to the majority of cellular users in the coming years [6,

7].

Time and Frequency being scarce resources, the impact and importance of scheduling is very

high in a LTE network. To work on such a topic will not only help me to understand the

present technology and solutions but also help develop a better solution for the future.

13

1.2.2 Thesis goals

The main purpose of this thesis is to verify and compare selected downlink and uplink

schedulers in LTE MAC Lab (aka Matlab version of 4G System Lab provided by IS-

Wireless) with their implementation in an LTE network test environment (deployed in the

Institute of Telecommunications within the Smart City of TPSA in the Warsaw University of

Technology). The simulation part of the thesis enables us to understand the scheduling

algorithms for the LTE networks in much more detail and gain experience in modelling and

simulation of such networks in detail. During this thesis a detailed study of the network

architecture and layers being proposed for LTE networks is carried out.

One of the main contributions of this dissertation is to work out a method to examine LTE

scheduling performance evaluation for teaching purposes.

1.3 Thesis Scope

This thesis is organized in 7 chapters. The rest of the chapters are organized as follows:

Chapter 2 gives an overview of LTE. Chapter 3 describes the concept of scheduling along

with the description of the scheduling algorithms under consideration. Chapter 4 discusses

the simulation and testing scenarios and results. Chapter 5 presents teaching proposals in the

form of a laboratory experiment. Finally chapter 6 draws the conclusion and gives

recommendations for future works. Chapter 7 details the various sources and references of

study.

14

2. An Overview of LTE

This chapter will provide an insight into the technical details of Long Term Evolution as

underlined by the 3GPP. The chapter starts with describing the LTE requirements, the

transmission schemes used for uplink and downlink, followed by other important features

like MIMO.

2.1 LTE requirements

The 3GPP has laid out specific requirements that need to be fulfilled by LTE which are listed

in [10], with some of them listed below:

Peak Data Rates:

E-UTRA should support significantly increased instantaneous peak data rates. The supported

peak data rate should scale according to size of the spectrum allocation.

Note that the peak data rates may depend on the numbers of transmit and receive antennas at

the UE. The targets for downlink (DL) and uplink (UL) peak data rates are specified in terms

of a reference UE configuration comprising:

a) DL capability 2 receive antennas at UE b) UL capability 1 transmit antenna at UE For this baseline configuration, the system should support an instantaneous downlink peak

data rate of 100Mb/s within a 20 MHz downlink spectrum allocation (5 bps/Hz) and an

instantaneous uplink peak data rate of 50Mb/s (2.5 bps/Hz) within a 20MHz uplink spectrum

allocation.

Latency:

A user plane latency of less than 5 ms one-way and a control plane transition time of less than

50 ms from dormant to active mode and less than 100 ms from idle to active mode.

User throughput:

Downlink:

2-3 times higher downlink throughput than High Speed Downlink Packet Access (HSDPA)

Release 6 at the 5% point of the Cumulative Distribution Function (C.D.F).

3-4 times higher average downlink throughput than HSDPA Release 6.

The user throughput should scale with the spectrum bandwidth.

Uplink:

2-3 times higher uplink than Release 6 Enhanced UL at the 5% point of the CDF.

2-3 times higher average uplink throughput than Release 6 Enhanced UL (EUL).

The user throughput should scale with the spectrum bandwidth provided that the Maximum

transmit power is also scaled.

Mobility:

LTE shall support mobility across the cellular network and should be optimized for 0 to 15

km/h. Furthermore, should support also higher performance at 15 and 120 km/h. Connection

15

shall be maintained at speeds from 120 km/h to 350 km/h (or even up to 500 km/h depending

on the frequency band).

Spectrum efficiency:

3-4 times higher spectrum efficiency (in bits/s/Hz/site) in downlink and 2-3 times higher in

uplink, compared to Release 6 HSDPA and EUL respectively.

Bandwidth/Spectrum flexibility:

LTE should support several different spectrum allocation sizes such as: 1.25 MHz, 1.6 MHz,

2.5 MHz, 5 MHz, 10 MHz, 15 MHz and 20 MHz. in both uplink and downlink where the

latter is used to achieve the highest peak data rate, with both TDD and FDD modes. It should

also support the flexibility to modify the radio resource allocation for broadcast transmission

according to specific demand or operators policy. Furthermore the communication can take place both in paired (FDD) and unpaired (TDD)

bands. Paired frequency bands means that the uplink and downlink transmissions use separate

frequency bands, while in the unpaired frequency bands downlink and uplink share the same

frequency band.

Coverage:

Cell ranges up to 5 km support the above targets; up to 30 km will suffer some degradation in

throughput and spectrum efficiency and up to 100 km will have overall performance

degradation.

Given some of the advantages of an OFDM approach, 3GPP has specified OFDMA as the

basis of its LTE effort.

2.2 Multiple Access Techniques

3GPP LTE have selected different transmission schemes in uplink and downlink due to

certain characteristics. OFDMA has been selected for downlink i.e. from eNodeB to UE and

SC-FDMA has been selected for uplink i.e. for transmission from UE to eNodeB [12].

2.2.1 Downlink - Orthogonal Frequency Division Multiple Access

(OFDMA)

For downlink transmission LTE uses OFDMA which splits the data stream into many slower

data streams that are transported over many carriers simultaneously. The main advantage of

many slow but parallel data streams is that it leads to elongation of the transmission steps

which in turn help to avoid the issues of multipath transmission on fast data streams. This

scheme helps allocate radio resources to multiple users based on frequency (subcarriers) and

time (symbols) using OFDM. For LTE, OFDM subcarriers are typically spaced at 15 kHz and

modulated with QPSK, 16-QAM, or 64-QAM modulation.

16

Fig. 1 : OFDM and OFDMA [28]

The full potential of OFDMA is utilised by proper scheduling as it allows the resources to be

used between multiple users flexibly by sharing the subcarriers, with differing bandwidth

available to each user versus time.

2.2.2 Uplink - Single Carrier - Frequency Division Multiple Access (SC-

FDMA)

For uplink transmission the use of OFDMA is not ideal because of its high PAPR when the

signals from multiple subcarriers are combined and hence as a result an alternative to OFDM

was sought for use in the LTE uplink. And as we know power consumption is a key

consideration for UE terminals and for this there was a need to adopt a transmission scheme

which wouldnt comprise with the requirements of LTE without putting too much pressure on the power consumption of UEs. The solution came up in the form of SC-FDMA that suits

very well with the LTE uplink requirements. The transmitter and receiver architecture is

nearly the same as OFDMA. Furthermore it also offers the same degree of multipath

protection.

Fig. 2 : OFDM and SC-FDMA [28]

In SC-FDMA instead of dividing the data stream and putting the resulting substreams directly

on the individual subcarriers, the time-based signal is converted to a frequency-based signal

with an FFT function. This distributes the information of each bit onto all subcarriers that will

be used for the transmission and thus reduces the power differences between the subcarriers

[18].

17

2.3 LTE Frame Structure

As per the description of LTE frame structure in [28] the downlink and uplink transmissions

are grouped in (radio) frame of length 10 milliseconds (ms). Each radio frame is divided into

10 subframes of 1ms duration each, with the subrame being further divided into 2 slots that

are 0.5 ms each. Each slot consists of 7 or 6 OFDM symbols for normal or extended cyclic

prefix used respectively [5]. The LTE frame structure is illustrated in the Fig. 3.

The smallest modulation structure in LTE is one symbol in time vs. one subcarrier in

frequency and is called a Resource Element (RE). Resource Elements are further aggregated

into Resource Blocks (RB), with the typical RB having dimensions of 7 symbols by 12

subcarriers. The RE and RB structure is also shown in Fig. 3. The number of symbols in a RB

depends on the Cyclic Prefix (CP) in use. During the use of normal CP the RB contains seven

symbols, whereas in case of extended CP which is used due for extreme delay spread or

multimedia broadcast modes, the RB contains six symbols.

Fig. 3 : LTE frame structure [18]

Due to the spectrum flexibility two frame types are defined for LTE, with Type 1 being used

in FDD while Type 2 is being used in TDD. Type 1 frames consist of 20 slots with slot

duration of 0.5 ms as discussed previously; whereas Type 2 frames contain two half frames,

where at least one of the half frames contains a special subframe carrying three fields of

switch information including Downlink Pilot Time Slot (DwPTS), Guard Period (GP) and

Uplink Pilot Time Slot (UpPTS). If the switch time is 10 ms, the switch information occurs

only in subframe one. If the switch time is 5 ms, the switch information occurs in both half

frames, first in subframe one, and again in subframe six. Subframes 0 and 5 and DwPTS are

18

always reserved for downlink transmission. UpPTS and the subframe immediately following

UpPTS are reserved for uplink transmission. Other subframes can be used for either uplink or

downlink. Frame Type 2 is illustrated in Fig. 4.

Fig. 4 : Frame Type 2 [27]

The number of RBs that can fit within a given channel bandwidth varies proportionally to the

bandwidth. Logically, as the channel bandwidth increases, the number of RBs can increase.

The transmission bandwidth configuration is the maximum number of Resource Blocks that

can fit within the channel bandwidth with some guard band [28]. The table 2.1 shows the

LTE bandwidth and resource configuration.

Table 2.1 : Bandwidth and Resource blocks specifications [1]

We can notice here that subcarrier spacing remains same in all bandwidth configurations. The

best results in terms of throughput can be achieved by the bandwidth with maximum amout

of RBs.

2.4 LTE Downlink Physical Channels

As in other networks like UMTS, all higher layer signalling and user data traffic are

organized by the means of proper channels. In LTE the downlink channels have been defined

in [28] the following way:

19

Fig. 5 : LTE Downlink channels [18]

We will be discussing the role and description of the physical downlink channels [28]

involved in LTE:

Physical Broadcast Channel (PBCH)

The PBCH is used to send cell-specific system identification and access control parameters

every 4th

frame (40 ms) using Quadrature Phase Shift Keying (QPSK) modulation. The

structure of PBCH is independent of the actual network bandwidth.

Physical Downlink Shared Channel (PDSCH)

The PDSCH is used to transport user data and is designed for high data rates. The Resource

Blocks associated with this channel are shared among users via OFDMA. The various options

for modulation include QPSK, 16- Quadrature Amplitude Modulation (QAM), and 64-QAM.

Spatial multiplexing is exclusive to the PDSCH.

Physical Control Format Indicator Channel (PCFICH)

The PCFICH is used to inform the UE how many OFDM symbols will be used for the control

information in PDCCH in a subframe. The number of symbols used ranges from 1 to 3. The

PCFICH uses QPSK modulation.

Physical Downlink Control Channel (PDCCH)

The PDCCH is used to inform UE about the uplink and downlink resource scheduling

allocations. It maps onto resource elements in up to the first three OFDM symbols in the first

slot of a subframe and uses QPSK modulation. The value of the PCFICH indicates the

number of symbols used for the PDCCH.

Physical Multicast Channel (PMCH)

The PMCH carries multimedia broadcast information with the use of modulation including

QPSK, 16-QAM, or 64-QAM. Multicast information can be sent to multiple UE

simultaneously.

20

Physical Hybrid Automatic Retransmission Request (HARQ) Indicator Channel

(PHICH)

PHICH carries Acknowledge (ACK)/Not Acknowledge (NACKs) in the downlink in

response to uplink transmissions in order to request retransmission or confirm the receipt of

blocks of data from the UE. ACKs and NACKs are implemented in HARQ mechanism.

Reference Signal (RS)

Reference signal is used by UE for downlink channel estimation. It allows the UE to

determine the channel impulse response (CIR). RSs are the product of a two-dimensional orthogonal sequence and a two-dimensional pseudo-random sequence. Since there are 3

different sequences available for the orthogonal sequence and 170 possible sequences for the

pseudorandom number (PRN), the specification identifies 510 RS sequences. The RS uses the

first and fifth symbols under normal CP operation, and the first and fourth symbols for

extended CP operation; the location of the RS on the subcarriers varies.

Primary and Secondary Synchronization Signal (P-SS and S-SS)

UEs use the Primary Synchronization Signal (P-SS) for timing and frequency acquisition

during cell search. The PSS carries part of the cell ID and provides slot timing

synchronization. It is transmitted on 62 of the reserved 72 subcarriers (6 Resource Blocks)

around Dedicated Control (DC) on symbol 6 in slot 0 and 10 and uses one of three Zadoff-

Chu sequences. UEs use the Secondary Synchronization Signal (S-SS) in cell search. It

provides frame timing synchronization and the remainder of the cell ID, and is transmitted on

62 of the reserved 72 subcarriers (6 Resource Blocks) around DC on symbol 5 in slot 0 and

10. The S-SS uses two 31-bit binary sequences and BPSK modulation.

2.5 LTE Uplink Physical Channels

In LTE the uplink channels have been defined in [28] the following way:

Fig. 6 : LTE Uplink Channels [18]

21

We will be discussing the role and description of the physical uplink channels [28] involved

in LTE:

Physical Uplink Control Channel (PUCCH)

The PUCCH carries uplink control information and is never transmitted simultaneously with

PUSCH data. PUCCH conveys control information including channel quality indication

(CQI), ACK/NACK responses of the UE to the HARQ mechanism, and uplink scheduling

requests.

Physical Uplink Shared Channel (PUSCH)

Uplink user data is carried by the PUSCH. Resources for the PUSCH are allocated on a sub-

frame basis by the UL scheduler. Subcarriers are allocated in units of RBs, and may be

hopped from sub-frame to sub-frame. The PUSCH may employ QPSK, 16-QAM, or 64-

QAM modulation.

Physical Random Access Channel (PRACH)

The PRACH carries the random access preamble and coordinates and transports random

requests for service from UEs. The PRACH channel transmits access requests (bursts) when a wireless device desires to access the LTE network (call origination or paging response).

Uplink Reference Signal

There are two variants of the UL reference signal. The demodulation reference signal

facilitates coherent demodulation, and is transmitted in the fourth SC-FDMA symbol of the

slot. A sounding reference signal is also used to facilitate frequency dependent scheduling.

Both variants of the UL reference signal use Constant Amplitude Zero Autocorrelation

(CAZAC) sequences.

2.6 Multiple Input Multiple Output

A major step in order to increase the data transmission rates was achieved by including

MIMO in the first release of LTE. LTE supports multiple antenna operation both in terms of

transmit diversity as well as spatial multiplexing with up to four layers. The use of MIMO

with OFDMA has some favourable properties compared to Wideband Code Division

Multiple Access (WCDMA) because of its ability to cope effectively with multi-path

interference [8]. The main features of MIMO help in improving the network performance as,

transmit diversity can be used to increase the robustness of communication in fading channels

by transmitting multiple replicas of the transmitted signal from different antennas whereas

spatial multiplexing can help increase the peak data rates as compared to the non-MIMO

scenarios by a factor of 2 to 4, depending on the eNodeB and the UE antenna configuration.

22

Fig. 7 : Single user MIMO transmission principle [8]

In LTE all the device categories with exception of the simplest, support MIMO capability.

Fig. 8 : Multi-user MIMO transmission principle [8]

The eNodeBs can also have multiple antennas without any adverse impact on the non-MIMO

UE as all devices can cope with the transmit diversity up to four antennas.

23

3. Selected Issues of Scheduling

This chapter along with explaining the concept of scheduling will give details of the various

downlink and uplink scheduling algorithms used in LTE.

In general terms, scheduling is basically the process of making decisions by a scheduler

regarding the assignment of various resources (time and frequency) in a telecommunications

system between users. In LTE the scheduling is carried out at eNodeB by dynamic packet

scheduler (PS) which decides upon allotment of resources to various users under its coverage

as well as transmission parameters including modulation and coding scheme (MCS). As

earlier discussed, LTE defines 1 ms subframes as the Transmission Time Interval (TTI)

resulting in the scheduling of resources every 1 ms. It means after every 1 ms the assignment

of resources could change depending upon various factors including CQI (Channel Quality

Indicator) sent as a feedback by the user to the eNodeB.

The process of selecting the transmission parameters and Modulation and Coding Scheme

(MCS) is known as Link Adaption (LA). Link adaption along with scheduling of resources is

meant to maximize the cell capacity, while making sure that the minimum Quality of Service

(QoS) requirements are met and there are adequate resources also for best-effort bearers with

no strict QoS requirements [8]. LA adjusts the data rate with the help of Adaptive modulation

and coding (AMC) by matching the modulation and the channel coding scheme on resources

assigned by the scheduler. In situations with advantageous channel conditions, AMC selects a

higher modulation order and coding rate and vice versa.

Fig. 9 : Layer 2 functionalities for dynamic packet scheduling, link adaptation, and HARQ

Management [8]

24

In LTE networks, the role of resource scheduling is very important because great

performance gain can be achieved by properly observing the amount of radio resources

assigned to each user. As the 3GPP hasnt standardised any scheduling algorithm, we are free to choose and implement any algorithm that would meet the expected our QoS. While

choosing or designing a scheduling algorithm many factors such as expected QoS level, the

behaviour of data sources, and the channel status have to be kept in mind. The problem

becomes more complex in the presence of users with different requirements in term of

bandwidth, tolerance to delay, and reliability [13].

3.1 Selected Scheduling Algorithms

There are various scheduling methods that have been developed over time to enhance the

process of scheduling. But in this thesis, we shall be concentrating on particularly three

algorithms which have been implemented in the software environment provided for testing by

IS-Wireless.

Among them are:

Round Robin, Max SNIR, and Proportional Fair.

3.1.1 Round Robin Scheduling

This scheduling method is based on the idea of being fair in the long term by assigning equal

no. of Physical Resource Blocks (PRBs) to all active UEs. It operates by assigning the PRBs

to UEs in turn i.e one after another without taking into account their CQI. Hence the users are

equally scheduled. For e.g. If we have 4 users U1, U2, U3, U4 and PRBs, this algorithm will

assign the resources in the following manner: U1, U2, U3, U4, U1, and U2. It can be

illustrated by the following flow chart:

Fig. 10 : Flow Chart for Round Robin Algorithm

25

The main advantage of this kind of scheduling is the relative ease in its implementation

whereas the major disadvantage is the fact that it does not take into account user CQI

feedback, which may lead to lower and unequal throughput.

3.1.2 Max SNIR Scheduling

The Max SNIR scheduling algorithm assigns the PRBs to the UE with the highest CQI on

that RB obtained in the form of feedback from the UE. Hence, for this method to work

properly the UE must feedback the CQI to the eNodeB. This algorithm helps in improving

the user throughput by assigning the PRB to the UE with good channel quality as a result

enhancing its peak data performance. The scheduling process can be seen in flow chart:

Fig. 11 : Flow chart for Max SNIR algorithm

Max SNIR algorithm can increase the cell capacity at the expense of the fairness. In this

scheduling strategy, UEs located far from the eNodeB (i.e. cell-edge) are unlikely to be

scheduled.

26

3.1.3 Proportional Fair Scheduling

This algorithm assigns the PRBs to the UE with the best relative channel quality i.e. a

combination of CQI & level of fairness desired. There are various versions of PF algorithm

based on values it takes into account. Main goal of this algorithm is to achieve a balance

between Maximising the cell throughput and fairness, by letting all users to achieve a

minimum QoS (Quality of Service).

Fig. 12 : Flow chart for Proportional Fair Algorithm

The above Fig. 12 depicts one of the possible methods of implementing proportional fair

algorithm. Such an algorithm is designed to be better in terms of average user throughput as

well as being fair to most of the users and meeting the minimum QoS requirements during the

scheduling process.

27

4. Simulations and Testing

This chapter is meant for the analysis of the scheduling algorithms we have discussed in the

earlier section by means of simulations and practical experiments. The analysis has been

carried out by comparing the throughput for different scenarios (different scheduling

schemes, different environment models and different number of users). Along with the

simulations, the practical work will be carried out in LTE network test environment

(deployed in the Institute of Telecommunications within the Smart City of TPSA in the

Warsaw University of Technology). This practical test environment is designed to analyse the

LTE network performance. A detailed description of each used tool is given in this chapter.

Then graphical representations of the performance of these scheduling algorithms in terms of

throughputs are plotted.

4.1 LTE MAC Lab System Level Simulator: An overview

According to the materials [24] provided by IS-Wireless: The Matlab version of 4G System

Lab also known as LTE MAC Lab belongs to the category of system-level simulators truly reflecting the behaviour of a modelled radio access network. A tool user selects and

configures the LTE radio interface, chooses appropriate channel and traffic models, defines

the network to be analysed, makes the choice on the set of Radio Resource Management

(RRM) functionalities and runs the simulation. The tool is time-driven and models, with high

accuracy, all the events that happen within a radio network. This includes terminals mobility,

cell selection / reselections, attach, random access, admission control, handovers, power

control, scheduling and many more. Special attention is given to co-channel interference

(CCI) control, where functionalities managing CCI can be easily modelled and verified. After

simulation, user-defined network statistics collected over defined observation time window

are available as reports for post processing.

The role played by RRM features in LTE, LTE-Advanced will be far greater than in previous

wireless systems. Therefore decision to put special emphasis on appropriate modelling of

RRM features was taken. This constitutes the cornerstone of 4G System Lab and differentiates it significantly from the classical RNP tools. LTE MAC Lab provides plenty of

representative algorithms for RRM and CCI control. It is a continuously evolving product. In

the near future it will include features belonging to 3GPP Rel 9 and more importantly to Rel

10 also known as LTE-Advanced.

4.1.1 Simulation Scenarios

In order to verify and compare the scheduling algorithms with the help of LTE MAC Lab, we

have selected no. of scenarios. These scenarios are meant to help us understand the working

of the scheduling algorithms in downlink and uplink. We investigate the performance of the

scheduling algorithms in terms of resource allocations and throughput for different scenarios

(different scheduling methods, different channel models and different number of users). Here

is a chart depicting the cases that are considered:

28

Fig. 13 : A tree diagram for all the scenarios under consideration for simulations

The scenarios have been selected to analyse the impact of the scheduling algorithms in

different conditions, hence understand their functioning in much more detail.

4.1.2 Simulation Results and Analysis

In this section all the simulation results are presented along with their analysis. During the

simulations we will be set some basic default parameters which are depicted below:

Table 4.1 summary of simulation parameters used for all the testing scenarios

Parameters Value

Number of Equipment (UEs) 1or 3

Number of base station 1

Bandwidth 3 MHz

Channel type Stationary and Vehicular (Highly Mobile)

Simulation length 150 TTI

Scheduling algorithms Round Robin, Max SNIR and Proportional Fair

Testing Scenarios

Uplink

Round Robin

.......

Max SNIR

.......

Proportional Fair

.......

Downlink

Round Robin

Single User Multiple

Users

Stationary

High Mobile

One user at cell edge

Max SNIR

......

Proportional Fair

........

29

Multipath Model 3GPP model

Environment Type Urban

Frequency 850 MHZ

Model Type 3GPP model

Base Station height 20 m

Base Station Antenna Characteristic Omnidirectional

User Equipment height 1.5 m

BLER 10^(-1)

FFT Size 256

Transmission Scheme SISO

The parameters have been considered so as to create the most appropriate simulation

environment that is relative to real scenarios.

Downlink Scenario

Case 1: Single User, High mobility, Using Round Robin, Max SNIR and Proportional Fair

scheduling algorithms

In this first case we simulate a single user and we show the resource allocations and user

throughput for different SNIR values. We have plotted graph depicting the SNIR measured

for each PRB which eventually impacts the scheduling along with a single graph depicting

the resource allocations and throughput, respectively, for different scheduling algorithms

(RR, Max SNIR, PF) as in case of single user the scheduling algorithm does not impact the

resource allocations as all resources are allocated by default to the single user.

Fig. 14 : PRB allocation based on SNIR values for single user downlink Case 1

0 2 4 6 8 10 12 14 160

5

10

15

20

25Measured SNIR for 1 user in Downlink in the frequency axis (3 MHz band)

PRB number

SN

IR in d

B

User 1

30

The SNIR values have been limited to the range of -2 dB to 25 dB like in a realistic scenario

so as to derive results which reflect the real life situation.

Fig. 15 : Resource Allocation for a single user in downlink Case 1

All the resources are assigned to the single user as no scheduling algorithm kicks in until

there is more than a single user under an eNodeB.

Fig. 16 : Throughput Results for single user in downlink Case 1

Fig. 15 depicts the allocation of resources to the user which in our case is complete allocation

for single user, followed by Fig. 16 which shows user throughput achieved. We observe that

the throughput is limited to around 4.5 5 Mbps because of various factors such as SNIR values. We can reach a Maximum throughput of 13.2 Mb/s, if all conditions are favourable.

20 40 60 80 100 120 140

2

4

6

8

10

12

14

16Scheduler allocations for single user

TTI

PR

B

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1

2

3

4

5

6Throughput vs TTI for 1 user

TTI * 10

Mbit/s

User 1

31

Case 2: 3 Users, Stationary, Using Round Robin, Max SNIR and Proportional Fair


In this case we simulate 3 users and we show the resource allocations and user throughput for

different SNIR values. We have plotted graph depicting the initial SNIR measured for each

PRB which eventually impacts all the scheduling algorithms followed by plots depicting the

resource allocations and throughput for different scheduling algorithms (RR, Max SNIR, PF).

Fig. 17 : PRB allocation based on SNIR values for 3 users

This figure is valid for all multi-user cases, as the initial SNIR settings will remain same

throughout our simulation experiments in order to compare data collected properly.

Fig. 18 : Resource allocation by RR algorithm for 3 users in downlink Case 2

0 2 4 6 8 10 12 14 160

5

10

15

20

25Measured SNIR for 3 users in Downlink in the frequency axis (3 MHz band)

PRB number

SN

IR in d

B

User 1

User 2

User 3

20 40 60 80 100 120 140

2

4

6

8

10

12

14

16Round Robin Scheduler allocations

TTI

PR

B

32

Fig. 18 depicts the allocation of resources by Round Robin algorithm in which each user gets

allocated the same number of resources.

Fig. 19 : Resource allocation by Max SNIR algorithm for 3 users in downlink Case 2

We can observe that the allocation of resources in Fig. 19 is as per the SNIR values of each

user; hence the higher a user has SNIR the more resources get allocated to it.


20 40 60 80 100 120 140

2

4

6

8

10

12

14

16MaxSNIR Scheduler allocations

TTI

PR

B

20 40 60 80 100 120 140

2

4

6

8

10

12

14

16Proportional Fair Scheduler allocations

TTI

PR

B

33

Fig. 20 shows how the PF algorithm first starts out by allocating equal no. of resources to

each user and then eventually shares the resources such that each user is able to attain highest

equal throughput.

Fig. 21 : Comparison of PRB allocation in all three algorithms over time Case 2

Fig. 22 : Comparison of throughput obtained from all three algorithms Case 2

In the above plots we can see the how each scheduling algorithm carries out the task of

resource allocation. In Fig. 18 we can see the resources being allocated in a cyclic way to

20 40 60 80 100 120 140

500

1000

1500MAX SNIR scheduler

TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

20 40 60 80 100 120 140

500

1000

1500Round Robin scheduler

TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

20 40 60 80 100 120 140

500

1000

1500Proportional Fair scheduler

TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1

2Throughput vs TTI for 3 users :Downlink Round Robin scheduler

TTI * 10

Mbit/s

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1

2Throughput vs TTI for 3 users :Downlink MAX SNIR scheduler

TTI * 10

Mbit/s

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1

2Throughput vs TTI for 3 users :Downlink Proportional Fair scheduler

TTI * 10

Mbit/s

User 1

User 2

User 3

34

each user irrespective of their SNIR values, in Fig. 19 we see the Max SNIR scheduler allots

more resources to the users having a higher SNIR than the other and in Fig. 20 we observe

the Proportional Fair scheduler assigning resources in terms of fairness in the beginning and

then trying to balance the fairness and best throughput results for each user. In Fig. 21 we can

analyse the allotment of PRB over time using each scheduling algorithm, it also shows the

fairness of these algorithms quite clearly. The last Fig. 22 shows the throughput results

achieved with the help of these scheduling algorithms and helps us compare them. We can

observe that Round Robin algorithm delivers fairness to all the users, the Max SNIR

algorithm has the Maximum throughput but not all users are able to enjoy the best speed and

the Proportional Fair algorithm tries to strike a balance between fairness and achieving the

Maximum throughput.

Case 3: 3 Users, High Mobile (Vehicular), Using Round Robin, Max SNIR and Proportional

Fair scheduling algorithms

In this case we simulate 3 users moving at a speed of 100 Kmph and we show the resource

allocations and user throughput for different SNIR values. We have plotted graph depicting

the initial SNIR measured for each PRB which eventually impacts all the scheduling

algorithms followed by plots depicting the resource allocations and throughput for different

scheduling algorithms (RR, Max SNIR, and PF).

Please refer to Fig. 17 for initial PRB allocations based on SNIR values.


Fig. 23 depicts the allocation of resources by Round Robin algorithm in which each user gets

allocated the same number of resources without taking into consideration any other

parameters.

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

35


We observe that the allocation of resources in Fig. 24 is still as per the SNIR values of each

user; the change of speed does not affect the allocations until unless the SNIR also changes

significantly, hence the higher a user has SNIR the more resources get allocated to it.

Fig. 25 : Resource allocation by PF algorithm for 3 users in downlink Case 3

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

36

We can notice the difference between the Fig. 20 and Fig. 25 both depicting the allocation as

per PF with the channel of channel conditions.

Fig. 26 : Comparison of PRB allocation in all three algorithms over time Case 3

Fig. 27 : Comparison of throughput obtained from all three algorithms Case 3

We can observe from Fig. 23 that Round Robin algorithm delivers fairness to all the users,

the Max SNIR algorithm has the Maximum throughput for user 2 but not all users are able to

enjoy the best throughput and the Proportional Fair algorithm tries to strike a balance

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1


TTI * 10

Mbit/s

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1

2


TTI * 10

Mbit/s

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1


TTI * 10

Mbit/s

User 1

User 2

User 3

37

between fairness and achieving the Maximum throughput. The throughput results are a bit

better due to the fact all users are located within 7-8 m radius from the base station and the

change of SNIR values has a positive effect in the small time frame under consideration, but

eventually as the distance of users from the eNodeB increases there should be degradation in

throughput.

Case 4: 3 Users (user 1 at cell edge), High Mobile (Vehicular), Using Round Robin, Max

SNIR and Proportional Fair scheduling algorithms

In this case we simulate 3 users moving at a speed of 100 Kmph with user 1 at cell edge and

we show the resource allocations and user throughput for different SNIR values. We have

plotted graph depicting the initial SNIR measured for each PRB which eventually impacts all

the scheduling algorithms followed by plots depicting the resource allocations and throughput

for different scheduling algorithms (RR, Max SNIR, and PF).



The changes in channel conditions still do not affect the behaviour of RR algorithm as it

allocates the resources equally among all users no matter what conditions.

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

38


Fig. 30 : Resource allocation by PF algorithm for 3 users in downlink Case 4

The above figure depicts the resource block allocation to the users. We observe that User 1 is

scheduled equally in the Round Robin algorithm in Fig. , but the Max SNIR algorithm shown

in Fig. 29, it has the least allocated resources due to its SNIR values. We also observe that in

the Proportional Fair algorithm depicted in Fig. 30 the algorithm tries to allot more resources

in order to facilitate higher throughput.

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

39

Fig. 31 : Comparison of PRB allocation in all three algorithms over time downlink Case 4

Fig. 32 : Comparison of throughput obtained from all three algorithms downlink Case 4

As a conclusion we can say that in the downlink the throughput results are not much affected

by the channel conditions provided that SNIR values do not change significantly. The Max

SNIR and RR scheduling algorithms provide consistent results while the PF algorithm tries to

achieve the best throughput within available conditions.

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1


TTI * 10

Mbit/s

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1

2


TTI * 10

Mbit/s

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1


TTI * 10

Mbit/s

User 1

User 2

User 3

40

Scenario Uplink

Case 1: Single user, Stationary, Using Round Robin, Max SNIR and Proportional Fair


In this case we simulate a single stationary user. We show the resource allocations and user

throughput for different SNIR values. We have plotted graph depicting the initial SNIR

measured for each PRB which eventually impacts all the scheduling algorithms followed by

plots depicting the resource allocations and throughput for different scheduling algorithms

(RR, Max SNIR, and PF).

Fig. 33 : PRB allocation based on SNIR values for single user in uplink Case 1

The same settings were chosen for the simulations in uplink also so as to facilitate analysis of

results.

0 2 4 6 8 10 12 14 160

5

10

15

20

25Measured SNIR for 1 user in Uplink in the frequency axis (3 MHz band)

PRB number

SN

IR in d

B

User 1

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

41

Fig. 34 : Resource Allocation for a single user in uplink Case 1

Fig. 35 : Throughput results of single user in uplink Case 1

Case 2: 3 users, Stationary, Using Round Robin, Max SNIR and Proportional Fair


In this case we simulate 3 stationary users. We show the resource allocations and user

throughput for different SNIR values. We have plotted graph depicting the initial SNIR

measured for each PRB which eventually impacts all the scheduling algorithms followed by

plots depicting the resource allocations and throughput for different scheduling algorithms

(RR, Max SNIR, and PF).

Fig. 36 : PRB allocation based on SNIR values for 3 users

This figure is valid for all multi-user cases, as the initial SNIR settings will remain same

throughout our simulation experiments in order to compare data collected properly.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5Throughput vs TTI for 1 user :Round Robin scheduler

TTI * 10

Mbit/s

User 1

0 2 4 6 8 10 12 14 160

5

10

15

20

25Measured SNIR for 3 users in Uplink in the frequency axis (3 MHz band)

PRB number

SN

IR in d

B

User 1

User 2

User 3

42

Fig. 37 : Resource allocation by RR algorithm for 3 users in uplink Case 2

We observe that the pattern of resource allocation in RR differs from downlink, this is due to

the use of SC-FDMA in the uplink.

Fig. 38 : Resource allocation by Max SNIR algorithm for 3 users in uplink Case 2

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

43

In the uplink the Max SNIR allocates as much consecutive PRBs as possible up to the point

when other user has better SNIR. In UL we cannot get non-consecutive allocations so all

PRBs must be consecutive.

Fig. 39 : Resource allocation by PF algorithm for 3 users in uplink Case 2

Fig. 40 : Comparison of PRB allocation in all three algorithms over time uplink Case 2

Fig. 40 helps us clearly see how the algorithms are allocating users over the whole simulation

period. From this we can also predict the results of the throughput, the more resources a user

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

44

gets the higher the throughput.

Fig. 41 : Comparison of throughput obtained from all three algorithms uplink Case 4

In the above plots we can see the how each scheduling algorithm carries out the task of

resource allocation in the uplink. We can also observe that the scheduling allocations in

downlink and uplink are not the same due to the fact that downlink uses OFDMA whereas

uplink uses SC-FDMA.

In Fig. 37 we can see the resources being allocated in a cyclic way to each user irrespective

of their SNIR values, in Fig. 38 we see the Max SNIR scheduler allocates as much

consecutive PRBs as possible up to the point when other user has better SNIR and in Fig. 39

we observe the Proportional Fair scheduler assigning resources in terms of fairness in the

beginning and then trying to balance the fairness and best throughput results for each user. In

Fig. 40 we can analyse the allotment of PRB over time using each scheduling algorithm, it

also shows the fairness of these algorithms quite clearly.

The last Fig. 41 shows the throughput results achieved with the help of these scheduling

algorithms and helps us compare them. We can observe that Round Robin algorithm delivers

fairness to all the users, the Max SNIR algorithm has the Maximum throughput but not all

users are able to enjoy the best speed and the Proportional Fair algorithm tries to strike a

balance between fairness and achieving the Maximum throughput.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1

2Throughput vs TTI for 3 users :Uplink Round Robin scheduler

TTI * 10

Mbit/s

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1

2

3Throughput vs TTI for 3 users :Uplink MAX SNIR scheduler

TTI * 10

Mbit/s

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

0.5

1

1.5Throughput vs TTI for 3 users :Uplink Proportional Fair scheduler

TTI * 10

Mbit/s

User 1

User 2

User 3

45

Case 3: 3 users, High Mobile (vehicular), Using Round Robin, Max SNIR and Proportional

Fair scheduling algorithms

In this case we simulate 3 highly mobile (100 Kmph) users. We show the resource allocations

and user throughput for different SNIR values. We have plotted graph depicting the initial

SNIR measured for each PRB which eventually impacts all the scheduling algorithms

followed by plots depicting the resource allocations and throughput for different scheduling

algorithms (RR, Max SNIR, and PF).




20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

46

In Fig. 43, we can notice that with the change in channel conditions the resource allocation

has been affected, this might be due to the SNIR change experienced by the different users.



In Fig. 44 we observe that the PF algorithm has to change how the resources are allocated so

as to accommodate the change in channel conditions.

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

47


Case 3: 3 users with User 2 at cell edge, High Mobile (vehicular), Using Round Robin, Max

SNIR and Proportional Fair scheduling algorithms

In this case we simulate 3 highly mobile (100 Kmph) users with user 2 at cell edge. We show

the resource allocations and user throughput for different SNIR values. We have plotted

graph depicting the initial SNIR measured for each PRB which eventually impacts all the

scheduling algorithms followed by plots depicting the resource allocations and throughput for

different scheduling algorithms (RR, Max SNIR, and PF).


1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1


TTI * 10

Mbit/s

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1

2


TTI * 10

Mbit/s

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

0.5

1

1.5Throughput vs TTI for 3 users :Uplink Proportional Fair scheduler

TTI * 10

Mbit/s

User 1

User 2

User 3

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PRB

48




We can notice in Fig. 48 and 49 that the change in channel condition of one user from the

previous state affects the change of allocation of resources significantly which was not the

case in downlink.

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

20 40 60 80 100 120 140

2

4

6

8

10

12

14


TTI

PR

B

49



In this case we observe how the user at the cell edge is unable to get resources allocated due

to its poor SNIR values. Even the Round Robin algorithm is not able to allocate resources to

this user due to the fact that in uplink, not all resources need to be allotted, but some can be

left blank, which hence results in lesser allocation of resources even in a fair algorithm like

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

20 40 60 80 100 120 140

500

1000


TTI

Num

ber

of

allo

cate

d P

RB

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1


TTI * 10

Mbit/s

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

1

2


TTI * 10

Mbit/s

User 1

User 2

User 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 150

0.5

1Throughput vs TTI for 3 users :Uplink Proportional Fair scheduler

TTI * 10

Mbit/s

User 1

User 2

User 3

50

Round Robin, Fig. 47. The proportional fair algorithm, Fig. 49 also tries to schedule user 2

but is unable to achieve the best results due to it conditions.

4.2 LTE network test environment

The LTE network test environment has been set up by the Institute of Telecommunications of

Warsaw University of Technology with the help of its partners. The testing environment

consists of a LTE eNodeB which is able to provide LTE coverage, along with UE meant to

test the network.

The main advantage of having such a testing environment is to be able to implement new

ideas and algorithms as well as analyse the data collected during the simulations with a

running model. It also helps find out practical challenges during network configuration and

operation which are not noticeable during the simulatio

Modeling and Simulation of Scheduling Algorithms in LTE Networks

Documents

Transcript of Modeling and Simulation of Scheduling Algorithms in LTE Networks