Extending the LTE-Sim for LTE-Advance with CoMP and Relaying in Heterogeneous 4G Mobile Networks

Transcription

1 University of Technology, Sydney Faculty of Engineering and Information Technology Extending the LTE-Sim for LTE-Advance with CoMP and Relaying in Heterogeneous 4G Mobile Networks Haider Al Kim Supervisor: Associate Professor Kumbesan Sandrasegaran The work contained in this report, other than that specifically attributed to another source, is that of the author(s). It is recognised that, should this declaration be found to be false, disciplinary action could be taken and the assignments of all students involved will be given zero marks. Signed: Date: 21/11/2014 i

2 Acknowledgement First of all, I would like to express my warm thanks to Imam Sahib Al Zaman (as) who is always a beacon shining on my way to success. This master thesis project is the final stage in obtaining the master degree in telecommunication networks at the University of Technology Sydney (UTS). This project was conducted in the Centre for Real-Time Information Networks (CRIN) in the faculty of engineering and information technology in the UTS. I have been working in this project from March 2014 to November During this project I have had much support from several people. I would like to express my honest gratitude below. Associate Professor Kumbesan Sandrasegaran has been my supervisor for this project. He was been a great support providing guidance, advice, constructive criticism and encouragement over the course of the last year. In addition, I am deeply and forever indebted to my parents. My sincere appreciation and gratitude to them is for their efforts and their distinctive role in all fields of my life, besides their faith in me and allowing me to be as ambitious as I wanted. Your prayer for me was what sustained me thus far. Importantly, my grateful thanks are extended to my wife, Ruwaida. Her support, encouragement, quiet patience and unwavering love were undeniably the bedrock upon which the past five years of my life have been built. Warm thanks for my brothers and sisters for their unwavering supports. Finally, for all of these people who motivated me to do the best and were confident that I will be the best, I offer this modest gift to express thanks. i

3 Acknowledgement... i Contents... ii List of Figures... iv List of Tables... v Abbreviation List... vi Abstract... x 1. Chapter 1: Introduction Background Motivation and goal of the project Motivation Thesis objective Thesis Scope Chapter 2: LTE-A Introduction LTE- Advance Enhancements Air Interface Enhancement Channel Bandwidth Structure Carrier Aggregation Effective and Guard bands Improving spectral efficiency Heterogeneous Network (HetNets) HetNets Challenges Higher Spectrum Utilization Signaling Optimizations Frequency Domain ICIC: Time Domain ICIC Network Based Techniques Advanced MIMO Scheme Transmission/Reception Coordinated Multi-Point Relays Summary Chapter 3: Radio Resource Management Introduction RRM in both DL and UL Connection Mobility Control (CMC) Handover Future Trends of Handover Handover Phases in LTE-A Admission Control Packet Scheduling (PS) Downlink Packet Scheduling Packet Scheduling Algorithms in Downlink Direction Uplink Packet Scheduling Power Control (PC) ii

4 Balancing of Carrier Load Carrier Load Balancing Interference Management Summary Chapter 4: LTE-Sim Heterogeneous Network Deployment Introduction Downlink System Model of LTE Packet Scheduling Algorithms Proportional Fair (PF) Algorithm Maximum Largest Weighted Delay First (MLWDF) Algorithm Exponential/Proportional Fair (EXP/PF) Algorithm Simulation.1- Single Macro Cell with two Pico Cells Simulation.1 Environment Simulation.1 Results Throughput Packet Loss Ratio (PLR) Delay Fairness Index Simulation.2-Single Macro Cell with two Pico Cells (Different Speed Comparison) Simulation.2 Environment Simulation.2 Results Throughput Packet Loss Ratio (PLR) Delay Fairness Index Simulation.3- Single Macro Cell with Increasing Pico Cells Simulation.3 Environment Simulation.3 Results Throughput Packet Loss Ratio (PLR) Delay Fairness Index Conclusion 91 iii

5 List of Figures Figure 2.1 Evolution of LTE-Advance... 4 Figure 2.2 Carrier Aggregation... 7 Figure 2.3 LTE-A Protocols Stack... 8 Figure 2.4 Aggregation Process... 9 Figure 2.5 Effective and Guard Bands with Aggregation Calculations Figure 2.6 Heterogeneous Network Example Figure 2.7 Driving Factors and enablers for small cell deployment Figure 2.8 Main Comparison between HetNets layers, MLC (Minimum Coupling Loss) Figure 2.9 Small Cell Extension concepts Usage to Offload Macro Cell Figure 2.10 CA-based ICIC in HetNets Figure 2.11 ABS concept to provide interference free in HetNets Figure 2.12 Flowchart indicate ABS information elements exchange over X Figure 2.13 SU-MIMO and MU-MIMO Figure 2.14 Advanced MIMO Figure 2.15 Coordinated Scheduling/Beamforming Figure 2.16 Joint Processing [28] Figure 2.17 Uplink Coordinated Scheduling Figure 2.18 Relays Node (RN) architecture Figure 2.19 Relays Duplexing Schemes Figure 2.20 FDD/TDD relay system Figure 2.21 A repeater protocol stack (layer 1 performing relaying) Figure 2.22 Layer 2 Protocol Stack (Decoding/Encoding) Figure 2.24 Protocol stack of RN Figure 2.23 Protocol stack (Layer 3) Figure 3.1 RRM functions and the mapping to the lower layers Figure 3.2 Principle of Macro Diversity Handover Figure 3.3 Principle of Fast Base Station Switching Handover Figure 3.4 Hard Handover Figure 3.5 Multicarrier Handover Figure 3.6 X2 Initiation Phase [34] Figure 3.7 X2 based Handover Preparation Phases Figure 3.8 S1 based Handover Preparation Phases Figure 3.9 Handover Execution Phase Figure 3.10 Handover Completion Phase-X1 based Handover Figure 3.11 Handover Completion Phase-S1 based Handover Figure 3.12 RRM Framework in LTE-A Figure 3.13 Interactions between HARQ, PS and LA Figure 3.14 Frequency DPS Concept Figure 3.15 Uplink RRM Functionalities inter-work with LA and PS Figure 3.16 enb Classification for LTE Rel 8 and LTE-A Arrival UEs Figure 4.1 An Example of HetNets Figure 4.2 Downlink Packet Scheduler of the 3GPP LTE System Figure 4.3 Applied HetNets (Macro with 2 Picos) Figure 4.4 Average System Throughput (Macro with 2 Picos) Figure 4.5 Average System Throughput (single Macro cell) Figure 4.6 PLR of Video Flows (single Macro cell) Figure 4.7 PLR of Video Flows (Macro with 2 Picos) Figure 4.8 Packet Delay of Video Flows (single Macro cell) iv

6 Figure 4.9 Packet Delay of Video Flows (Macro with 2 Picos) Figure 4.10 Fairness Index of Video Flows [15] Figure 4.11 Fairness Index of Video Flows Macro with 2 Picos Figure 4.12 Throughput of Video in Macro with 2 Picos Figure 4.13 PLR of Video in Macro with 2 Picos (3 Km/h and 120 Km/h speed) Figure 4.14 Delay of Video in Macro with 2 Picos (3 Km/h and 120 Km/h speed) Figure 4.15 Fairness Index in Macro with 2 Picos (3 Km/h and 120 Km/h speed) Figure Applied HetNets (Macro with Multiple Picos Scenarios) Figure 4.17 Throughput Gain of Video traffic in Macro with 2-10 Picos Scenarios Figure 4.18 Throughput Gain of Video traffic in Macro with 2-10 Picos Scenarios Figure 4.19 PLR Video traffic Comparison in Macro with 2-10 Picos Scenarios Figure 4.20 PLR of Video traffic in Macro with 2-10 Picos Scenarios Figure 4.21 Delay of Video traffic Comparison in Macro with 2-10 Picos Scenarios Figure 4.22 Comparison Delay of Video traffic in Macro with 2-10 Picos Scenarios Figure 4.23 Fairness Index in Macro with 2-10 Picos Scenarios Figure 4.24 Fairness Index in Macro with 2-10 Picos Scenarios List of Tables Table 2.1 LTE-A agreed requirements... 5 Table 2.2 Carrier Aggregation Models... 7 Table 3.1 QCI Parameters for EPS Bearer QoS Profile Table 4.1 Mapping between instantaneous downlink SNR and data rate Table 4.2 LTE System Simulation Parameters Table 4.3 Pico Cells Positions in meters into the Macro Cell (Radius 1 Km) Table 4.4 Throughput Gain Values and An Average of The Values Table 4.5 PF Throughput Gain Values and An Average of The Values v

7 Abbreviation List 1G First Generation 2G Second Generation 3G Third Generation 4G Fourth Generation 3GPP Third Generation Partnership Project 3GPP2 Third Generation Partnership Project 2 AC Admission Control ACK Acknowledgement AMBR Aggregate Maximum Bit Rate AMC Adaptive Modulation and Coding APFS Advanced Proportional Fair Scheduler ARP Allocation Retention Priority ARQ Automatic Repeat Request AS Access Stratum ATB Adaptive Transmission Bandwidth BM-SC Broadcast Multicast Service Centre BS Base Station CA Carrier Aggregation CC Carrier Component CCCH Common Control Channel CDMA Code Division Multiple Access CN Core Network CoMP Cooperative Multipoint Transmission and Reception CP Cyclic Prefix CQI Channel Quality Indicator CRC Cyclic Redundancy Check CRS Cell specific Reference Signal CS/CB Coordinated Scheduling/Beamforming CSI Channel State Information CSI-RS Channel State Information Reference Signal DCCH Dedicated Control Channel DFT Discrete Fourier Transform DL Downlink DM-RS Demodulation Reference signal DRA Dynamic Resource Allocation DTCH Dedicated Traffic Channel DwPTS Downlink Pilot Time Slot EDGE Enhanced Data Rates for GSM Evolution EHR Efficient HARQ Retransmission enb Evolved Node Base station EPC Evolved Packet Core EPF Enhanced Proportional Fair vi

8 EPS Evolved Packet System E-UTRAN Evolved UMTS Terrestrial Radio Access Network EV-DO Evolved Data Only EV-DV Evolved Data Voice FBSS Fast Base Station Switching FDD Frequency Division Duplex FDMA Frequency Division Multiple Access FDPS Frequency Domain Packet Scheduling FFR Fractional Frequency Reuse FFT Fast Fourier Transform FRF Frequency Reuse Factor FSHO Fractional Soft Handover GBR Guaranteed Bit Rate GP Guard Period GPRS Generalized Packet Radio System GSM Global System for Mobile communication HARQ Hybrid Automatic Repeat Request HAS HARQ Aware Scheduling HHO Hard Handover HOL Head-Of-Line HSDPA High Speed Downlink Packet Access HSS Home Subscriber Service ICI Inter Cell Interference ICIC Inter cell Interference Coordination IDFT Inverse Discrete Fourier Transform IFFT Inverse Fast Fourier Transform IMT 2000 International Mobile telecommunication 2000 IS 95 Interim Standard 95 IMT-Advanced International Mobile Telecommunication Advanced ITU-R International Telecommunication Union Radio-communication JP Joint Processing LA Link Adaptation LTE Long Term Evolution LTE-A Long Term Evolution Advanced MAC Medium Access Control MBMS Multimedia Broadcast Multicast Channel MBMSGW MBMS Gateway MBR Maximum Bit Rate MBSFN Multimedia Single Frequency Network MCCH Multicast Control Channel MCE Multi-cell/Multicast Coordination Entity MDHO Macro Diversity Handover MH Mobile Hashing vii

9 MIMO MISO M-LWDF MME MTCH MU-MIMO NACK NAS OFDM OFDMA OLLA PARP PBCH PC PCCH PCFICH PCRF PDCCH PDCP PDSCH PF PFS P-GW PHICH PHY PMCH PMI PRACH PRB PS PSD PUCCH PUSCH QCI QoS RAN RAPF RAS RB RE RLC RN ROHC RR Multiple Input Multiple Output Multiple Input Single Output Modified-Largest Weighted Delay First Mobility Management Entity Multicast Traffic Channel Multi User Multiple Input Multiple Output Negative Acknowledgement Non Access Stratum Orthogonal Frequency Division Multiplexing Orthogonal Frequency Division Multiple Access Outer Loop Link Adaptation Peak-to-Average Power Ratio Physical Broadcast Channel Power Control Paging Control Channel Physical Control Format Indicator Channel Policy Charging Rule Function Physical Downlink Control Channel Packet Data Convergence Protocol Physical Downlink Shared Channel Proportional Fair Proportional Fair Scheduling Packet Data Network Gateway Physical HARQ Indicator Channel Physical Layer Physical Multicast Channel Precoding Matrix Indicator Physical Random Access Channel Physical Resource Block Packet Scheduling Power Spectral Density Physical Uplink Control Channel Physical Uplink Shared Channel QoS Class Identifier Quality of Service Radio Access Network Retransmission Aware Proportional Fair Retransmission Aware Scheduling Resource Block Resource Element Radio Link Control Relay Node Robust Header Compression Round Robin viii

10 RRM RRU RSRP SAE SC-FDMA SFR S-GW SIMO SINR SIR SISO SHO SRS SSDT SSHO SU-MIMO TB TDD TDMA TPC TSN TTI UE UL UMTS UpPTS WCDMA Radio Resource Management Radio Resource Unit Reference Symbol Received Power System Architecture Evolution Single Carrier Frequency Division Multiple Access Soft Frequency Reuse Serving Gateway Single Input Multiple Output Signal to Interference plus Noise Ratio Signal to Interference Ratio Single Input Single Output Soft Handover Sounding Reference Signal Site Selection Diversity Transmission Semi Soft Handover Single User Multiple Input Multiple Output Transmission Block Time Division Duplex Time Division Multiple Access Transmit Power Control Time Sequence Number Transmission Time Interval User Equipment Uplink Universal Mobile Telecommunication System Uplink Pilot Time Slot Wideband Code Division Multiple Access ix

11 Abstract This report presents heterogeneous network (HetNets) in the Long Term Evolution (LTE) to introduce Long Term Evolution-Advanced (LTE-A). The evolution in the next generation of mobile network has been stated in this study using the Pico with Macro HetNets. Such networks are under what is so-called 4G technology that meets users aspirations in terms of data rate and system accessibility. LTE and LTE-A provide high speed access to the packet data rate; therefore, various devices such as notebook, IPods, smart phones, laptops, and cameras currently could be connected to the internet to work in their full features. Most recent networks depend on the functionality of enhanced base station to perform the complex operations; thereby, rely on Radio Resource Management (RRM) functionalities that is placed in enhanced Node B. RRM is demonstrated focusing on its functions such as packet scheduling and handover management. Taking the advantage of HetNets while utilizing of LTE-based operations such as Carrier Aggregation (CA), Multi-in Multi-out antenna MIMO and Cooperation Multipoint transmission and reception CoMP has been widely adopted by mobile operators since the cost of HetNets (adding small cells) is considerably accepted. This mixing of HetNets with LTE specific technologies improves spectral efficiency, enhances the system coverage and capacity, as well as minimizes the overall cost of the operating. More importantly, it is expected that it boosts the data rate to 1 Gbps in the downlink direction and 500 Mbps in the uplink direction and supports a speed of mobility up to 500 Km/h. The Third Generation Partnership Project (3GPP) target was obtaining 100 Mbps high peak data rate in the downlink and 50 Mbps in the uplink using the 20 MHz bandwidth of LTE system comparing with the previous systems. Due to the limited available radio resources, RRM performs packet scheduling to allocate resource fairly among instantaneous arrived users. The system performance is affected by the packet schedulers that play an essential role in the resource allocations. This study is based on three selected packet scheduling schemes that have been built in the used simulation platform. Real Time algorithms such Maximum-Largest Weighted Delay first (M-LWDF) algorithm and the exponential/proportional fair (EXP/PF) have been implemented. The Non-Real Time algorithm that is used is Proportional Fair (PF). The performance of these schemes is evaluated via the metric of the throughput, Packet Loss Ratio PLR (also called Packet Error Rate), delay (latency) and fairness index. x

12 1. Chapter 1: Introduction 1.1. Background The mobile telecommunication systems have been developed since 1980s. The first generation 1 G started the domination on the mobile market using the analogue scheme besides Frequency Division Multiple Access (FDMA) technology. The features of 1G involve consuming of high power and using narrow frequency bands; therefore, 1G was ineffective system. The second generation 2G came to overcome the drawbacks of 1G; as a result to the revolution in the digitized cellular networks. For example, Interim Standard 95 (IS-95) and Global System for Mobile communication (GSM) are second generation mobile schemes. Qualcomm, an American company, designed IS-95 as a mobile technology in USA. IS-95 was built based on the technique of Code Division Multiple Access (CDMA) to support maximum bit rate 14.4 Kbps. In the early 1987, Europe initially proposed GSM to provide roaming service. Later, since the use of harmonized spectrum, the international roaming can be applied throughout the globe and hence GSM is accepted by various countries. It allows to the subscribers to be served from most of the places on the plant that operate GSM using the same mobile number. GSM was based on circuitswitched network for voice call only, but later the data services are added to the system. The technology that was utilized by GSM was Time Division Multiple Access (TDMA) and the maximum bit rate that could be reached with GSM was 9.6 Kbps. While the revolution was continuing in the wireless networks, more enhancements for both IS-95 and GSM were introduced. These developments emerged to support more bit rate and utilize of the available spectrums efficiently. IS-95B was the enhanced IS-95 to while Generalized Packet Radio System (GPRS) are included in GSM to support data services since the GSM as aforementioned was developed initially to voice service. Further improvements to GSM system were done to introduce what is well-known as Enhanced Data Rates for Global Evolution (EDGE). IS-95, GPRS and EDGE are under the 2.5G. In the late of 1990s, Third Generation Partnership Project (3GPP) which is a united group of telecommunications standard organizations defined the third generation (3G). The 3G was based on the Wideband CDMA (WCDMA) technology that provides 5 MHz wideband of CDMA besides supporting a frequency reuse operation of 1. Another feature of WCDMA was the data rates integration on a single carrier using the flexible physical layer. In theory, the data rate of 1

13 WCDMA should be 2 Mbps. On the other hand, Third-Generation Partnership Project 2 (3GPP2) standardized mobile technologies in USA; thereby, cdma2000 was the evolved IS-95B. Video on demand, video conferencing and mobile TV are real-time applications that use 3G networks [1]. 3GPP and 3GPP2 launched High-Speed Downlink Packet Access (HSDPA) and cdma Evolved Data Only (1 EV-DO) respectively in beginning of These technologies are classified under 3.5G, which contain new enhancement methods for the mobile network such as Hybrid Automatic Repeat Request (HARQ), distributed architecture, scheduling operation and modulation and coding schemes (MCS) [2]. Six years later, IEEE released the Worldwide Interoperability for Microwave Access (WiMAX) that was standardized as IEEE e. WiMAX competed HSDPA and EV-DO technologies offering high data rate and better spectral efficiency. It relied on Orthogonal Frequency Division Multiplexing (OFDM) as its access technology. The Long Term Evolution (LTE) of the Universal Mobile Telecommunication System (UMTS) has been developed as a consequence of the demand for a competitive technology in order to satisfy users experiences. The main goals of LTE system are enhancing the performance, increasing capacity and coverage and reducing delay time and deployment cost while maintaining the simplicity of the network. Using 20 MHz of bandwidth, LTE was planned to support maximum bit rate of 100 Mbps /50 Mbps in the downlink /uplink respectively. Moreover, the latency of the user plane was decided to be reduced to less than 5 ms while the delay of the control plane was aimed to be less than100 ms. 350 Km/h was proposed as the speed of mobility for LTE users and 100 Km as a coverage area for LTE network. 3GPP website ( has the full LTE requirements and features for detailed information. More recently, the advanced LTE, also called Release 10, have taken the attention of the network operators. LTE-A is an enhanced system of LTE that is anticipated to surpass LTE. The planned features of LTE-A are mainly introducing higher bit rate (up to 1 Gbps in the downlink and 500 Mbps in the uplink) and attaining higher speed of mobility (500 Km/h). Rel 10 (LTE-A) has adopted number of new technologies in order to achieve that. These technologies involve: heterogeneous networks (Macro with Pico, Femto and relaying), Carrier Aggregation (CA), CoMP and advanced MIMO scheme. 2

14 1.2. Motivation and goal of the project Motivation The encouragement to do this project arises from the demand to investigate the performance of LTE-A which is expected to dominate the future mobile networks. More users will be switched to LTE and LTE-A as predicted 80% of mobile broadband users in the near future. Due to the fact that the operation cost should be minimized and the utilization of the available radio resources should be as efficient as possible, Radio Resource Management (RRM) is considered the key tool that has to be focused on to be improved. It has the functions that can be configured to improve the current telecommunication networks. The trade-off between deploying RRM functionalities is the main goal of investigating these mechanisms in order to obtain more reliable system, higher throughput besides lower transmission delay. Since applying heterogeneous networks is cost effective method to improve the LTE, the focus is on HetNets Thesis objective This study has mixed between investigating the current LTE system performance and introducing the LTE-A by deploying heterogeneous networks. The first purpose has been achieved by investigating one of the main RRM functions that is packet scheduling in the downlink direction. The well-known scheduling algorithms; Proportional Fair (PF) algorithm, Maximum Largest Weighted Delay First (MLWDF) algorithm and Exponential Proportional Fair (EXP/PF) algorithm, have been used. An open source simulation platform called LTE-Sim has been utilized that includes these algorithms. The second purpose is to develop a new code within LTE-Sim platform that could be considered an extending to the current LTE-Sim to create LTE- A environment in order to investigate LTE-A system performance. This integrated code is a scenario of HetNets (Macro with Pico cells) using the aforementioned scheduling schemes. The system using these algorithms is examined based on the metrics of Packet throughput, Packet Error Rate, packet latency (delay), and fairness index Thesis Scope The thesis is organized as follows. Chapter 1 gives a historical overview and then states the motivation and the objectives. Chapter 2 focuses on the HetNets and LTE-A besides LTE technologies in general. Chapter 3 explores the main functions of RRM in both LTE and LTE-A focusing on handover and packet scheduling. Chapter 4 is the technical papers of this thesis and in the end a proposal for a doctoral study are included. 3

15 2. Chapter 2: LTE-A 2.1. Introduction Long Term Evolution (LTE) was evolved to ensure that its technology satisfy the International Telecommunication Union Recommendation requirements by using International Mobile Telecommunication 2000 project (IMT-2000) of the ITU-R. This development ensures that the LTE remains competitive for predictable future needs. LTE Rel-8 requirements are enhancing system coverage and capacity, improving user experience by providing higher data rate and lower latency. Moreover, decreasing cost of operation and deployment and seamless backward compatibility are other LTE demands. LTE has to meet with the IMT-advanced, therefore; further improvements were conducted in These improvements involve: firstly, data rate increment from 100 Mbps up to 1 Gbps in downlink (DL) direction and from 75 Mbps up to 500 Mbps in the uplink (UL) direction. Secondly, spectral efficiency increment utilizing 8 8 antenna layout in the DL direction to get 30 bps/hz and using 4 4 antenna layout in the UL direction to get 15 bps/hz. Thirdly, declining latency of control plane in changeover from camped and dormant to active state to be 50 ms and 10 ms respectively [11].Summarized Table 2.1 shows the LTE-A required requirements. Several advancements have been proposed in order to reach these demands in the network deployment and system performance, thereby, introducing the LTE-A network. These improvements are involving carrier aggregation, advanced MIMO including beamforming with spatial multiplexing enhancement in the UL/DL directions, relay nodes deployment and transmission/reception cooperation multipoint CoMP. In this chapter, these technologies are discussed. Figure.2.1 shows LTE-A development and number of technologies and applications applied in release 8, 9 and 10 of LTE. Figure 2.1 Evolution of LTE-Advance [11] 4

16 Items Maximum data rate Requirements 10 Gbps Downlink direction 500 Mbps Uplink direction Maximum spectral efficiency User spectral efficiency in Cell-edge User spectral efficiency in Average Cell Latency of Control Plane 30 bps/hz (MIMO 8x8 ) Downlink direction 15 bps/hz (MIMO 4x4 ) Uplink direction 0.12 bps/hz (MIMO 4x4 ) Downlink direction 0.07 bps/hz (MIMO 2x4 ) Uplink direction 3.7 bps/hz (MIMO 4x4 ) Downlink direction 2 bps/hz (MIMO 2x4 ) Uplink direction 50 ms (Camped Active state) 10 ms (Dormant Active state) Latency of User Plane Lower than Rel 8 Table 2.1 LTE-A agreed requirements [2] 2.2. LTE- Advance Enhancements It could be classified the main enhancements of LTE-A compare with LTE as the following aspects: Air Interface Enhancement Channel Bandwidth Structure In LTE Rel8/9, the total bandwidth is (20 MHz) represents one carrier component (CC). In LTE- A using the heterogeneous networks where cells are overlapped, carrier aggregation can be applied. It allows to multiple small bandwidth segments called carrier components to create wider virtual frequency band in order to transmit at higher rates. The standard number of aggregated CCs to represent 100 MHz of LTE-A bandwidth is five component carriers. This is used to achieve 1 Gbps/500 Mbps in DL/UL directions. On the other hand, it offers backward compatibility to LTE users, in which the LTE users can only use one component carrier (20 MHz) while the LTE-A users utilize up to 5 components carrier to achieve LTE-A users 5

17 requirements. However, no all the bands are available to be allocated to LTE-A users. This is because the CC has two parts: effective band and guard band. The effective part consists of the physical radio blocks (PRB) which is the efficient part of the band that can be allocated to the subscribers [30] Carrier Aggregation In LTE-A (Release 10), carrier aggregation (CA) has been introduced for providing bandwidth extension up to 100 MHz by aggregating multiple 20 MHz carrier components (CCs). It maintains a compatibility with LTE releases 8 and 9 while increasing the required bandwidth to meet LTE-A requirements. This increment in the bandwidth will increase the data rate in LTE-A significantly to provide a peak up to1 Gbps (downlink) and 500 Mbps (uplink). Each CCs has two parts: effective band and gap band. Effective bandwidth is equal to the total contiguous physical radio block (PRB) times the total bandwidth subtracting the gab band (GP) [30]. Equation 1.1 illiterates the effective bandwidth that used in CA of each CC. Effective BW = (1- GB%) x PRB [30] (2.1) In general, CA could be classified into three sorts un the term of the mechanism in which frequencies of CCs are companied as shown in Figure.2.2 [11]: Intra-band aggregation, contiguous component carriers: duplexing mode is FDD or TDD. While FDD allows asymmetric CA to get larger bandwidth in DL than UL, TDD provides symmetric CA since the same carriers has been used in DL and UL. However, it is possible to TDD to provide asymmetric CA using various time splits in downlink and uplink [2]. Intra-band aggregation, non-contiguous component carriers: FDD or TDD is the duplexing mode. Inter-band aggregation, non-contiguous component carriers of different frequency band (multi-band). The duplexing mode is FDD or TDD. Table 2.2 provides more details about all possible scenarios. 6

18 Figure 2.2 Carrier Aggregation [29] In the advanced LTE, 3GPP differentiated four implemented models for carrier aggregation, as illustrated in Table 2.2 [29]. These models comprise both non-contiguous multiple frequency bands CA using FDD and TDD modes and contiguous single frequency bands using FDD and TDD modes. Models Carrier Aggregation Deployment model A Uplink: 3.5 GHz - 2x20 MHz Downlink: 3.5 GHz - 4x20 MHz FDD contiguous allocation: single band (Uplink: 40 MHz, Downlink 80 MHz) B Uplink: 2.3 GHz - 5x20 MHz TDD contiguous allocation: single band (100 MHz) C D FDD non-contiguous allocation: multi band for (Uplink: 30 MHz, Downlink: 30 MHz) TDD non-contiguous allocation: multi band (90 MHz) Table 2.2 Carrier Aggregation Models [20] 7

19 There is another group of spectrum bands provided by 3GPP in addition to aforementioned LTE- A carrier aggregation spectrums; these spectrums are [20]: The GHz bands The 3,4 3.6 GHZ and GHz bands The MHz bands The MHz bands The MHz bands The GHz bands The GHz bands There is a similarity between LTE Rel-8 and LTE-A protocol architecture, the LTE Rel-8 control plane architecture is applied to CA of LTE-A. However, in the user plane; the LTE-A has a difference in which PDCP and RLC layers cannot see CCs operation. On the other hand, HARQ of each CC in the MAC layer handle to the physical layer in the DL direction or from the physical layer in the UL direction. Figure.2.3 shows protocols stack for LTE-A [20]. Figure 2.3 LTE-A Protocols Stack [20] 8

20 Effective and Guard bands There are different algorithms to aggregate the carrier components in the intra-band and the more complicated algorithms that applied in the inter-band. The procedure that responsible for allocation component uses the Effective bands to be allocated to the LTE-A users. The Effective bands are the actual affordable bands that can be used to be allocated to the requested user in LTE-A. This leaves a gap to separate between these effective bands, which is called Guard band. Guard bands are mainly used to avoid Doppler Effect for high mobility users. While the orthogonality is used to avoid the interference between carrier components, this Doppler Effect causes non-negligible impact on the orthogonality between frequency bands in LTE. As mentioned before, there are actual bands that can be allocated which mean that it cannot allocate all available bands. The following equation is used to calculate the total available bandwidth (resource) to be allocated the LTE-A users. Guard band (GB) and PRB is Physical Resource Blocks (consisted of subcarriers, the smallest elements that used to carry user data) Effective Bands = (1- GB%) x PRB bandwidth. To generalize the allocation procedure, the following diagrams shows that Figure 2.4 Aggregation Process 9

21 In LTE, CA has supported only 5 CC each one with 20 MHz. Not all the bandwidth of 20 MHz is available to be aggregated due to the gap band (Guard band). Hence, the total band that is used to be allocated to LTE or LTE-A users can be calculated. The following Figure.2.5 illustrates the concept of effective band, the guard band and the aggregation process. is the channel bandwidth, is the subcarrier bandwidth, is the contiguous subcarriers and is the total percentage of guard band (GB) [30]. Figure 2.5 Effective and Guard Bands with Aggregation Calculations [30] Improving spectral efficiency Among different base stations, the same carrier frequency (co-channel deployment) is shared. In addition, support of localized high traffic-densities ( hot-spots ) and deliver an increase in capacity simultaneously. The main improvement is using the Heterogeneous Network (HetNets). 10

22 Heterogeneous Network (HetNets) The motivation factor for HetNets is that there are considerable technological and economic causes for the rapid deployment of heterogeneous networks. The results of this technological enhancement are expected to have profound effects on the future telecommunication. Normally, any mobile operator installs new base stations to cope the increasing of traffic demand, choosing the transmission power and antenna configuration in order to complement the existing cells. This combination of large and small cells will lead to co-exist various Radio Access Technologies (RATs). Generally, HetNets can be defined as a mix of macro cells,low power cells such as (Femto cells,pico cells, and relays), and remote radio heads (RRH) with multiple RATs (Figure.2.6), to bring the network closer to the end users and increase the user expectation. The main reason behind adaptation of HetNets in the recent telecommunication (LTE-A) is that the radio link performance, theoretically, has been reached its limitations. Hence, logically the next performance jump must come from the diversity of wireless technologies. The main driving factors to use small cells in turn create HetNets are illustrated in Figure.2.7.However, one of the main challenges of applying HetNets is the intra-frequency interference [34]. One the other hand, the measurement that is used to differentiate between base station classes (how close the user to the base station) is Minimum Cabling Loss (MCL). If it is more that 70 db, the base station type is a wide range (macro cell over 300 meters coverage). When it is 53 db, it is medium range base station (micro cell, meters coverage). For local area ones, the MCL is 45 db which means that the user is very close to the base station ( Femto or Pico cells, less than 50 meters coverage) [34]. Figure 2.6 Heterogeneous Network Example 11

23 Figure 2.7 Driving Factors and enablers for small cell deployment [34] Mainly, there are five types (layers) of cells which can construct HetNets. The below explains each sort: A- Macro cell: it has a wide antenna to provide coverage for several square kilometers, utilizing high power transmission and high mounted antennas [34]. B- Micro cell: it is outdoor antennas that are smaller than macro cells. It covers only a few hundreds of meters by using low antennas deployments [34]. C- Pico cell: is a class of small cells, could be referred to as an enterprise femto cell or metro femto cell (more details of femto in D). It reuses all available radio resource (it is called Co-channel deployment) that is used by larger cells (macrocellular network) to serve as an expansion of a macro cell [34]. Compare to the other class of small cells (femto), this class usually has more subscribers. Moreover, it provides data and voice services in larger promises than femto such as indoor in place coverage, for example, shopping center or outdoor hotspot coverage for instant a busy shopping street. Pico cells are designed to be environmentally hardened to be deployed outdoor, perfectly installed with enhanced antennas. Unlike femto cells that are specified to be used by only the members of the closed subscribers group, pico cells can be used by all qualified users. However, it has been noted that there may not a huge different between femto and pico with regard to the number of users and transmission power [36]. 12

24 D- Femto Cell: is the other class of small cells, and it is used as indoor cell only. The indoor solutions can be placed in any building, shopping center, office or even at home, and it is connected to macro cell base station using indoor antennas and RF cables [34]. It could be defined that the Femto cell is a small cell that has Home evolved Node B (HeNB) in order to provide UEs with the connections to a mobile operator s network, for instance, domestic IP broadband connections. It has low power capability; hence, the coverage of HeNB is small, thus, the cell size is small. Low power femto cells can be interpreted to lower cost equipments. This means motivation of scalability ubiquitous utilization. It is considered, from the operator point of view due to the cost-efficient, the means of capacity development and coverage expansion. The first standard-base Femto cell release was enabled in Rel 8 that can be deployed in any vendor due to a number of agreed 3GPP specifications. From the user viewpoint, there is no Femto since the operators provide a high level of connectivity and services. It offers better connection to the mobile network. On the other hand, it is used to offload the Macro cell providing enhanced service to the mobile terminal [33]. The main reason behind using Femto cells is that it improves the coverage and capacity in small promises such as home or small office [34]. Figure.2.8 shows a general comparison between the four main heterogeneous networks layers. E- Relay: a network repeater that less cost than deploying new cell (more details in relay section in this chapter later). Figure 2.8 Main Comparison between HetNets layers, MLC (Minimum Coupling Loss) [34] 13

25 HetNets Challenges While adoption heterogeneous network in the modern telecommunications networks has a significant impact of future network utilization and satisfies users expectations, there are considerable challenges encounter deploying it. Power disparity issue between large cells (macro) and small cells (pico and femto) is one of the main problems. This comes through different coverage area of macro cells and pico/femto cells. To solve this issue and to steer more traffic toward small cells, range cell extension has been introduced which virtually increases the size of a small cell. This can be done through the basic biasing; that means, the UE that receives stronger signal from macro cells would be forced to connect to small cell nevertheless [34].It is very effective and simple method to increase small cells offload. However, it should be noted that using CRE should be with high care since it could be lead to problematic interferences situations. Figure.2.9 illustrates the concept of range cell extension (CRE). The another obstacle faces using HetNets is that the co-channel interference problems which means use same radio resources by the operator for both small cells and large cells causing interference where the macro cell receives interference from pico or femto cells and vies versa. The proposed solution to solve the interference issue in LTE-A HetNets is that using what well known with enhanced time domain inter-cell interference coordination (eicic) with ABS (Almost Blank Subframes) [32]. There is another type of (eicic) based on carrier aggregation as elaborated in [35] which called enhanced CA-based ICIC with cross carrier scheduling. Figure 2.9 Small Cell Extension concepts Usage to Offload Macro Cell 14

26 Higher Spectrum Utilization. Compare to LTE, the advanced LTE that is comprised from HetNets has higher spectrum utilization. By companying multiple carrier components in LTE-A, so any effective PRB can be aggregated to create the bandwidth (BW) for LTE-A subscribers. The other strategy to increase the utilization of spectrum in LTE-A is that by using Statically Multiplexing (STM) method, in which the user can utilize of any resource blocks as long as they are affordable to be assigned, then after finishing its transmission, it will release them. This means more efficient than static allocation [30] Signaling Optimizations As it is mentioned before, there are two types of mechanisms to manage the problem of inter-cell interference between HetNets proposed by 3GPP [36]. Carrier aggregation based ICIC (frequency domain) and ABS (time domain ICIC) are these methods Frequency Domain ICIC - Carrier Aggregation based ICIC: enb can apply cross carrier scheduling if CA is supported. It could be applicable when the enb controls both the victim cell and the aggressor cell. For example, the victim cell is the pico cell and the aggressor cell is the macro cell. In such a scenario, enb can use cross-carrier scheduling to avoid using PDCCHs on the same carrier frequency. Details explanation of this scenario can be accessed in [36]. Figure.2.10 shows CAbased ICIC of 2 separate component carriers, different network layers at a time are assigned the primary component carrier (f1) and the second component carrier (f2). The f1 can be used by the macro layer to schedule its control information. However, it can still schedule its users on both f1 and f2. In turn, the interference on control and data can be avoided by scheduling control and data information for both macro and pico layers on different component carriers. As shown in the third subframe in Figure.2.10, it is also possible to schedule data information of users in Pico enb on the same carrier that the Macro layer schedules its users, as the interference from the aggressor cell (macro) on pico users can be tolerated. In contrast, pico UEs in the range extension region are still scheduled in the other carrier where UEs of the macro are not 15

27 scheduled. One backward of CA with cross carrier scheduling is that only Rel 10 and onwards users can be supported so this feature cannot be used by the old releases (8/9) [37]. Figure 2.10 CA-based ICIC in HetNets [37] Time Domain ICIC In this mechanism, the subframes are partitioned into two sets to be used by the HetNets layers, the victim cells use one set while the other set is used by the aggressor cells. Certain subframes used by the aggressor cell have to be muted by avoiding scheduling on those subframes so the victim cell can use these subframes to scheduling its UEs. The interfering cell avoids using the traffic channel during these blanked subframes. However, it still sends some essential information and signaling. This muted subframe is called almost blank subframe (ABS) that will be explained in more details in the following section [36]. - Almost Blank Subframes (ABS) Because the muted subframes are still carrying the signaling and other information, these empty subframes are called Almost Blank Subframes [36].It is almost blank to offer backward compatibility with Rel8/9. In other words, ABS are subframes with decreased transmission power including no transmission on some physical channels in the downlink direction. It is a form of blanking time; macro cell does not be allowed to transmit during it. ABS can be used by the victim cell (small cell) allowing cell range extension (CRE) UEs to get high quality signal and transmit with better conditions [32]. The UEs that suffer from a high level of interference should be served during these blanked subframes. In contrast, the users who are nearer to the transmitting enb that have not been impacted by interference can be served during the normal subframes (co-channeled subframes) [37]. Figure.2.11 shows ABS concept. 16

28 As mentioned before, the Almost Blank Subframes are designed to continue sending signaling and information. These signals are: - Cell Specific Reference Signal (CRS); - Acquisition channels for such as paging and broadcast, i.e. PSS/SSS/PBCH/SIB1/Paging/PRS [37]. Figure 2.11 ABS concept to provide interference free in HetNets [37] The pico users have been classified to two sets in term of ABS: 1- Cell Range Extension UEs: the users who suffer from high level of interference caused by macro enb should be serviced during ABS where the interference at its minimum value [37]. 2- Center Pico Cell UEs: the users who are closer to the center of pico cell are not highly impacted by macro cell interference since they maintain a good channel quality from their serving enb. As a result, the center pico users can be served with any subframes during ABS or non-abs [37]. 17

29 - ABS Information Elements Exchange It is a number of ABS bits used to assist the interfered cell with its process of scheduling. The victim enb is informed by each bitmap of ABS about the aggressing cell intention of power level. ABS bitmap can be divided into two types: aggressor cell bitmap and victim cell bitmap. The former has two main bitmaps which they are: A- ABS Pattern Info: represents ABS bit to aid interfered (pico) cell with its scheduling decisions. It is the first bitmap of ABS used to indicate which subframes the interfering (macro) cell has configured to be ABS [32]. B- Measurement Subset: it is obvious from its name that they are a set of subframes used for measurements objectives by the UEs of the victim enb [32]. The later has three main bitmaps which they are: A- Invoke Indication IE: requesting the ABS pattern from the aggressor enb. B- Usable ABS Pattern Info IE: is it intended to inform the sending enb the ABS subframes that are utilized by the receiving node [32]. C- DL ABS Status IE: indicates the ABS resource utilization status at the victim enb to the aggressor enb [32].Figure.2.12 below shows the ABS information elements exchange. Figure 2.12 Flowchart indicate ABS information elements exchange over X2 18

30 Network Based Techniques Advanced MIMO Scheme LTE Rel-8 uses (4 4 SU-MIMO) in which four antennas to the same user are dedicated in the downlink transmission and only one antenna in the uplink transmission. In Rel8, there are also four main various downlink transmission data modes: UE-specific RS-based beamforming, Multi-user MIMO, Open/Close loop spatial multiplexing and Open loop transmit diversity which have the following mode numbers ( 7; 6; (3,4&5); 2) respectively. Essentially, two streams of data for a single user (SU-MIMO) or two users get the same stream of data simultaneously (MU-MIMO).Mainly is geared towards TDD using Dedicated Reference Signals (DRS). Figure.2.13 shows the SU-MIMO and MU-MIMO. Figure 2.13 SU-MIMO and MU-MIMO 19

31 In LTE-A, MIMO operations are enhanced using new Rel10; mode number (9), in downlink and uplink transmission. Before mode 9, mode 8 used in Rel9 is introduced that is mainly geared toward TDD because the spatial multiplexing at the base station is got hold of using the sounding reference signal (SRS) [39]. In the downlink direction, spatial multiplexing is developed to support 8 8 antenna configuration to improve the performance and obtain eight data stream. This results in higher peak data rate in which double value is reached over Rel8. In addition, an evolved reference signal is deployed to assist a number of beamforming schemes. In the uplink direction, on the other hand, the baseline is 2 2 MIMO antenna design while 4 4 could be applied to provide peak data rate and improve the performance at the cell edge [28]. Recently, the concept of Massive MIMO has been proposed in LTE-A by means of 3.5 GHz. It is possible to align the individual antenna elements very close to each other. This enables the use of several tens or hundreds of antenna elements together. In Massive MIMO, beamforming is used with narrow beams. This reduces the interference and improves signal quality at cell edge because energy will be concentrated in a small area. However, MIMO has many problems that have to be addressed before the use of it in operational networks [31]. Figure.2.14 shows the advanced MIMO Figure 2.14 Advanced MIMO: 20

32 Transmission/Reception Coordinated Multi-Point Cooperative Multi-point or what so-called (CoMP) is a scheme for coordination among diverse number of enbs. These enbs are geographically separated and are linked via high speed dedicated connection elements, such as microwave links or fiber optics links. The purpose of CoMP is to enhance the users and system performance in the cooperation region[28]. As X2 is the interface that connecting the enbs in LTE-A, it will be used for performing CoMP process. While the number of coordinating enbs increases, the performance is getting better. The intercell interference effect in both the uplink and downlink directions is mitigated using an affirmative technique of coordination between enbs. In the downlink direction, coordinated transmission among enbs can be conducted in which two operations have been proposed: Coordinated Scheduling/Beamforming (CS/CB) and Joint Processing (JP).However, in the uplink direction, coordinated reception among enbs can relieve the interference. The only scheme that has been applied is coordinated scheduling in this direction, below is a description of each approach: Coordinated Scheduling/Beamforming (CS/CB). One enb takes the responsibility to transmit data to the users. However, a group of enbs shares control information specifically scheduling/beamforming decision as shown in Figure Figure 2.15 Coordinated Scheduling/Beamforming [28] 21

33 Joint Processing (JP). In this scheme, to remove interference and enhance received signal strength, several coordinated transmitting nodes altogether transmit data to the served UE. There are two ways to perform JP: fast cell selection and joint transmission. In the fast cell selection approach, the data is transmitted by one of the enbs at a time as shown in the right side of Figure In contrast, in the joint transmission approach enbs participate simultaneously to send data to the served terminal as shown in the left side of Figure However, it is considered a waste of resources since multiple enbs serve a single UE. This is because the signal power from some enbs may be feeble. In the JP operation, the coordinated base stations are served one UE to increase the micro-diversity. The operation depends on the CSI. Ideally, if CSI is available to the base station in its optimum value for all channels, cooperated antennas of all base stations can create a mechanism look like traditional MIMO. This can help of decreasing, and managing interference occurred between the UEs signals by using zero-forcing beamforming or multi-user MIMO techniques (MU-MIMO) or MMSE. The aim of using MU-MIMO is to get near-optimum performance. However, it is highly sensitive to the CSI accuracy. Because JP relays on CSI feedbacks, any delay in X2 interface can outdate CSI and lead to inefficient JP. On the other hand, any delay of CSI can be reduced by exchanging CSI through the air. However, this results in increasing the overhead and causes the difficulty to manage the interference that comes through these control messages. acts as a single base station In Comparison to JP, CS/CM, from its detention perspective, seems as a single base station serves the UE. It is more effective because the other enbs that are participating in the operation require less CSI messages between them. While CS/CS depends on the cooperated cells to avoid the interference, it ignores the received traffics from other enbs in the system considering them as pure interference. However, the diversity and multiplexing gain in CS/CM is less than JP since the UE served by one base station only [32]. 22

34 Figure 2.16 Joint Processing [28] Coordinated scheduling approach. Several geographically separated base stations corporate together by receiving the transmitted signal from UEs to increase cell-edge user throughput as shown in Figure.2.17.Compare with aforementioned approaches, this is used in the uplink direction. Figure 2.17 Uplink Coordinated Scheduling [28] 23

35 - CoMP s challenges CoMP approaches encounter the following obstacles: Extensive Overhead: could be per aggregated feedback or point feedback. It is a trade-off between overhead, delay and accuracy. Specific Reference signals of UE: UEs are unaware of the detailed operation in the network. Hence, some sorts of reference signals still may be required. Capacity of X2 interface or backhaul: massive information messages are needed for CoMP implementing depending on low-latency and high-bandwidth X 2 interfaces. Overload of the control channel: two main corporation types Joint Proportional and Coordinated Scheduling/Beamforming. Joint operation leads to increasing the number of UEs that their scheduling is conducted in the same subframe. Due to capacity of recent PDCCH, this could restrict the performance of scheduling that is required with CS/CB and JP. [32] Relays Relaying in LTE-A is another technique that is used in order to reduce the update of existing LTE system. The major consideration of designing the relay node is to expand the cell coverage area of LTE network. The Relay Node is a cost effective which is a cheap approach to providing coverage for far regions where the quality is poor or no service [38]. In addition, high data rate, throughput at the cell edge, temporary deployment of network and group mobility can be achieved by implementing relays in LTE-A [22]. Figure.2.18 shows the basic relays representation architecture in LTE-A. There are two main interfaces in relay system: Uu and Un. Uu interface is used to communicate the UE with the Relay Node (RN). Unique in LTE relay system, new interface known as Un is introduced which is used in connection between a donor enb and relay node. 24

36 Figure 2.18 Relays Node (RN) architecture [28] Terminology of LTE Relay: Relay Node (RN): repeater station. RN Cell: the area (Cell) that is covered by Relay Node. User Equipment (UE): term that is equivalent to MS (mobile station) in GSM system. It represents the end user terminal in LTE system. Donor enb (DeNB): the base station in the LTE architecture is called evolved node B (enb). If enb supports relay in LTE, it is called Donor enb. Donor enb Cell is the cell in LTE in which the relay functionality is supported by the base station (DeNB). Uu is the interface in which the user equipment can access the radio network. In relay scenario, it is a link between the relay node (RN) and the UE. In LTE relay system, it is also known as access link. Un is the connector between the RN and the DeNB. In LTE relay system, it is called backhaul link. 25

37 The user terminal is unaware whether it is connected to RN or enb. An RN, from the terminal perspective, looks like a normal cell in LTE system. This is; the data transfer and the signaling messages are the same with the case of non-relay cell. However, the security level in relays adds new challenges for system design of the relays. This is because the relay is new intermediate part of LTE network. Similar to LTE-A enb, RNs in LTE-A are required to be compatible with LTE UEs [38]. On the other hand, there are two frequency bands (inband and outband) utilized in the connection between RN and enb. If the frequency band used in the connection between UE and enb is the same that connect between RN and donor cell. This type is used to reduce the complexity. In contrast, the outband means various frequency bands are used in the link between RN & enb and enb & UE. - Deployment Scenarios of Relay Node (RN) There are various scenarios that have been defined since the relay is deemed better than a normal enb installation. The following summarizes the identified scenarios: A- Extension of Coverage: at a cell edge, a relay (RN) can be deployed to be used as an extension of the coverage for the enb. A normal deployment of RN would be in the rural regions at the cell edge where less population is present [36]. B- Reduction of dead spots: in a dead spot in which a coverage hole exists, a RN can be implemented to overcome it. The main reason behind existence a coverage hole is the physical obstruction, for example tunnel, building and so on so forth [36]. C- Enhancement of throughput: to boost the throughput in a particular are such as an indoor area or a hot spot, a relay (RN) can be deployed [36]. D- Temporary Coverage: when special events such as sport games and music courts are held, a relay node can be deployed to offer reliable services for the UEs in the hosting area which normally would be crowded [36]. E- Group Mobility: it is possible to deploy relay node (RN) in transportation such as a train, bus, and so on. Compare with other aforementioned cases, the relay node in this case is subjected to the mobility [36]. 26

38 - Duplexing Schemes Either TDD or FDD can be used by RN to connect with the UE and enb. Essentially, while TDD is a half-duplex communication, FDD is a full duplex communication. The following are the basic duplexing modes adopted to use spectrum resources in communications between network elements of LTE-A (UE, RN and enb): In the DL transmission between DeNB to RN and RN to UE, a basic TDD relay happens in 1 and 2 timeslots respectively. However, in the UL transmission, connection between UE and DeNB through RN happens in the next timeslots (3 and 4) respectively. Figure (a) illustrates that. Both in downlink and uplink directions, a basic FDD relay requires pair of frequency bands along with two time s slots as shown in Figure.2.19 (b). UE, RN and DeNB are communicated in the UL and DL simultaneously utilizing various orthogonal frequencies to avoid traditional inter-cell interference and interference between relay links (backhaul and access).the inband relay system is considered, so the same frequency is used in Un and Uu. Such system is so-called extended FDD relay and shown in Figure.2.19 (c). Figure 2.19 Relays Duplexing Schemes [28] 27

39 - Inband Relay As aforementioned previously, there are two sorts of inband relay system: the FDD and TDD. Due to the additional interference that relay suffers from which is due to the use of the same frequency in access and backhaul, the relay is designed to have subframe with nonoverlapping time zone. In the uplink and downlink, a pair of carriers is used with a time gap to separate backhaul link and access link. UE is unaware about these guard times and should connect to RN normally. The approach that is used to avoid the confusion and keep the backward compatibility is MBMS (Multimedia Broadcast Multicast Service) configuration. In this method, the relay system deludes the UE that the unused time zone as a useful MBSFN (Multicast Broadcast Single Frequency Network) subframe. This subframe is mainly used to provide MBMS in LTE [38]. Figure 2.20 FDD/TDD relay system In general, the mechanism used to connect the RN to the DeNB is adopted from the method that a UE connects to the enb. The same protocol stack with some modifications over the UE protocol stack is used. To keep backward compatibility, RN operates as enb to serve UE. Hence, the physical layer channel design has no significant difference in relay system. However, the backhaul link (Un) has modifications to meet the relay operations. New physical channels have been developed to meet the requirement of relay operation in the backhaul side of the relay network. Relay is similar to the conventional LTE physical signaling channel and data transmission channels (in DL and UL).It has similar PDCCH (physical download control channel) which is called R-PDCCH (relay physical download control channel), and 28

40 PDSCH/PUSCH (physical downlink shared channel/ physical uplink shared channel) which called R- PDSCH/R-PUSCH. Relay can be categorized with regard to various characteristics. This classification can be according to its functionality at each layer, duplexing types or according to the frequencies used in communication in Un and Uu links [28]. - Layers The classification of relay nodes can be conducted depending on which layers they work in. A repeater or what so called a layer 1 RN is responsible for amplification the arrived signal, then forwarding it to another network element which is another RN or a UE in the telecommunication network of the heterogeneous network. It is normally as a repeater amplifies any useful signal it receives as well as the undesirable signals such as noise and interference. This fact implies that it is used only in an environment where a high SNR. The layer 1 relay has a main advantage which is very fast method to forward the received signal. That is, it could be interpreted as a small delay appeared as furthermore multipath to the UE since no data passing over to the upper layers to be handled. It works in the physical layer, in which a donor enb RRC controls it [29, 36]. Figure.2.21 shows layer 1. Figure 2.21 A repeater protocol stack (layer 1 performing relaying) [36] 29

41 Decoding and Forwarding Relay: a layer 2 (L2) relay. It is responsible for decoding and re-encoding the arriving traffic before retransmitting it to the required UE. Unlike layer 1 relay node, it chooses only the desirable signal to amplify it. For this reason, it can be applied in a low SNR situation. Although the processing time is increased slightly, the layer 2 decoding and encoding process can override noise and interference. Since RLC and MAC layers are below the layer 2 relay type, it performs the upper layer functions of radio resource management such as data formatting, scheduling, and retransmission. As layer 1, it has to be controlled by DeNB since there is no RRC in the RN [29, 36]. Figure.2.22 shows layer 2. Figure 2.22 Layer 2 Protocol Stack (Decoding/Encoding) [36] RRC layer: a layer 3 (L3) relay. The similarity between layer 3 and layer 2 is that the noise and interference can be discarded by the processing of relay node L2. However, it is unlike layer 2 since it is capable of performing full L3 functions. Moreover, it has its own RRC together with layer 1 and layer 2 capabilities. Hence, it can control its cells without the need to DeNB RRC with their PCIs apparently to the UEs as a conventional enb element. RRC layer RN can be deemed as a wireless enb backhaul which is the disadvantage of this layer. Obviously, more signaling overhead and high efficiency are required in the wireless connection in this scenario, in turn, this increases the processing delay [29, 36]. Figure.2.23 shows layer 3 RN. 30

42 Figure 2.23 protocol stack (Layer 3) [36] - Radio Interface Protocol Stack of Relay Network Figure 2.24 Protocol stack of RN 31

43 As in traditional LTE, two main interfaces connect the relay system components. They are X2 and S1. While X2 interface is used to connect the donor enb (DeNB) to another enb in the network, S1 interface connects the far core network to the donor enb. However, the relay system use proxy term referring to donor enb that used as a proxy for RN, more specifically, X1 Proxy and S1 Proxy architecture are used [36] Summary To sum up, the first chapter discusses the detailed improvements on Rel-8 network to create LTE-A environment, in which considerable requirements such as data rate increment, delay time reduction and cell edge performance issue have been discussed. This chapter also deals with the challenges that confront the development LTE and. A detailed description of technologies that are adopted by LTE-A has also been proposed in this chapter. These technologies are mainly carrier aggregation, HetNets, advanced MIMO antennas, transmission and reception coordination multipoint, and the relay node. It should be noted that this chapter has detailed explanations for a specific issues such as inter-cell interference and the proposed solutions that overcome them as instance ABS operation. 32

44 References [28] I. F. Akyildiz, D. M. Gutierrez-Estevez, and E. C. Reyes, "The evolution to 4G cellular systems: LTE-Advanced," Physical Communication, vol. 3, pp , [29] AL-Jaradat, Huthaifa 2013, Radio Resource Management in LTE and LTE-A [30] Zhang, R.; Zheng, Z. ; Wang, M. ; Shen, X. (Sherman); Xie, L., 'Equivalent Capacity Analysis of LTE-Advanced Systems With Carrier Aggregation', pp [31] Korhonen, J. 2014, 'Introduction to 4G Communications', pp [32] Su, T.; Pang, J.; Su, HJ. Jun 2012, 'LTE-Advanced Heterogeneous Networks: Release 10 and Beyond', pp [33] SeungJune Yi, S.C., YoungDae Lee, SungJun Park, SungHoon Jung 2012, Radio Protocols for LTE and LTE-Advanced [34] Holma H, Toskala A 2012, LTE-Advanced 3GPP Solution for IMT-Advanced [35] Hu, Rose Qingyang Qian, Yi 2013, Heterogeneous Cellular Networks (2nd Edition). [36] Yi, Seunglune Chun, SungDuck Lee, YoungDae 2012, Radio Protocols for LTE and LTE- Advanced. [37] Shaer, H.E. 2012, 'Interference Management in LTE-Advanced Heterogeneous Networks Using Almost Blank Subframes'. [38] Dixit, H.-Y.W.J.R.S., WiFi, WiMAX, AND LTEMULTI-HOP MESH NETWORKS: Basic communication protocols and Application Areas WILEY. [39] A. Ghosh and R. Ratasuk, Essentials of lte and lte-a: Cambridge University Press,

45 3. Chapter 3: Radio Resource Management 3.1. Introduction In the recent telecommunication networks, an important and new tool called Radio Resource Management (RRM) has been used. The increments of the required services with a high level of transmission reliability and throughput, as well as the minimum level of delay, are the main reasons behind using the RRM. It is not only the aforementioned reasons, but also the radio elements are decreasing due to the increasing of users demands. In general to achieve the maximum resource utilization, RRM is using the affordable adaptation approaches such as link adaptation, users scheduling and hyper automatic repeat request (or so called HARQ). On the other hand, RRM manages the users according to their QoS requirements that have been agreed by both users and the networks providers. In the Figure 3.1, the RRM functionality and the mapping process from RRM to the various lower layers factions are shown. It also shows the control plane and users plan at the enhanced node B (enb). Mainly, the factions of RRM are classified into two sorts: semi-automatic and fully automatic. The former functions are performed at the third layer when a data flow is started, for example, admission control and permanent scheduling and management of QoS. Unlike semi-automatic, the fully-automatic functions are conducted the lower layers (1 and 2) at each new transmission time period which is normally 1 ms. Examples of such functions are link adaptation (LA), H-ARQ and scheduling of packets [3]. Figure 3.1 RRM functions and the mapping to the lower layers [3] 34

46 The network element that is responsible for the RRM functions in the LTE and LTE-A is the enhanced node B (enb) due to the distributed network architecture and removing the functions of Radio Network Controller (RNC). However, basic reports and information are still required such as Channel State Information (so-called CSI) in order to guarantee the best utilization of resources. These are the resources that can be allocated to the UEs by the resource allocator according to the status of the channel RRM in both DL and UL The Radio Resource Management is a collection of methods and algorithms that manage telecommunication system elements such as frequency, power, and modulation/coding. It ensures that the users get the agreed QoS while utilization from the finite affordable radio resources as efficient as possible. In the uplink and downlink, the main functions of RRM are similar. However, there are some limitations encounters each direction that can be detailed separately. The following are the main strategies that used in RRM: Connection Mobility Control (CMC) In the Radio Resource Control RRC, there are two main mobility modes of connection, idle mode and connection mode.cmc is responsible for the managing the radio resources in the (RRC IDLE) or (RRC CONNECTED) in which the connection parameters are set. The threshold and hysteresis are the parameters that used in the idle mode to enable the users from defining a cell or re-selecting new cell using reselection algorithms. More complexity has to be applied in the connection mode in which the resources mobility has to be presented (i.e., Handover).The enb and UE feedbacks and reports can be used to measure the required handover decision. However, more parameters could be utilized to take this decision such as the load in the adjacent cells, the predefined-policies of the operator and the traffic allotment. On one hand, it should be noted that in the idle mode the handover is made explicitly by the UE even though there is information provided by the network about cell selection and reselection. On the other hand, the mobility of UE in the connected mode is made by enb with or without measurements and reports from the UE to take the handover decision as mentioned previously. 35

47 Handover Handover could be defined as the operation in which new radio link is created between the serving enb (so-called source enb) and new target enb to hand the active UE to the better receiving signal enb. In general, there are two different sorts of handover; within the wireless system technology is call intra-handover such as handover in the LTE network between base stations and with other wireless communication systems called inter-handover. An example of inter-handover is the one that occurred between GSM and LTE. Further classification of handover whether intra or inter is that soft handover (SHO) and hard handover (HHO). The soft handover is shown in the legacy system such as GSM where the UE creates a new connection if the single strength is better with the target base station before leaving the serving one (source BS). This rule is a well-known as make before break. The soft handover supports the data to be delivered to the UE simultaneously from more than BS. Although the soft handover algorithms are more complex than hard handover, it provides smoother handover and reduces the probability of outage[28]. SHO has two main techniques in wireless telecommunication networks explaining in the following: 1- Macro Diversity Handover (MDHO): In this technique, there are a set of BSs called active set or diversity set that the UE can connect. While data is sent from all the BSs in the diversity set to the UE in the downlink direction, in the uplink direction, all active group BSs are responsible for receiving and processing data sent by the UE. In the system, there are also adjacent BSs for the active group. These BSs are monitored by the UE and can receive UE signals. However; the signal strength is insufficient to add the neighboring BSs to the active list. MDHO provides seamless, fast and stable handover which in turn reaches the system to a better performance. The drawback of this method is that the complexity in term of its algorithm and handover procedure compare with the hard handover. As a consequence, it is also considered that it wastes network resources and increases the system overhead due to the parallel synchronization between BSs from one side and between UE and BSs from the other side. MDHO is applied in UMTS and WiMAX [29].Figure 3.2 shows the principle of Macro Diversity Handover (MDHO). 36

48 Figure 3.2 Principle of Macro Diversity Handover [29] 2- Fast Base Station Switching (FBSS): Similarly, to MDHO, the UE connect to the group of BSs known as the active set. The different in this technique is that the UE monitors all BSs in the diversity set and decides; considering the signal strength, one of them as the anchor BS [29]. The UE is capable of connecting to only the anchor BS in the active list BSs for all downlink and uplink exchanges including control messages. For this reason, it is obvious that the overhead will be reduced using FBSS comparing with MDHO. In addition, the smoother traffic transfer from the serving base station to the receiving base station is supported using this type of handover. However, FBSS suffers higher data lost latency and higher outage probability in comparison to MDHO. Figure 3.3 illustrates the principle of this handover. Figure 3.3 Principle of Fast Base Station Switching Handover [29] 37

49 Hard handover, on the other hand, is based on the rule break before make that means that the UE is connected to the target enb after breaking up its connection with the serving enb. In E- UTRAN, only one cell is always serving the UE, in turn, the soft handover is not supported because it needs more than a single connection simultaneously to make the handover operation. For this reason, only hard handover is used in LTE system [33]. If the signal strength of the target enb is higher than the original signal received from the source enb by the UE, the UE is hard handed over. Figure 3.4 illustrates the HHO. Figure 3.4 Hard Handover As aforementioned before, LTE uses the only hard handover which has some drawbacks that have to be addressed. The following methods have been adopted in LTE to overcome the HHO shortages. 1- Semi-Soft Handover Mechanism (SSHO): this technique is adopted based on the macro diversity mechanism (MDHO). It is a mix of hard handover and soft handover, so it utilizes the advantages of both. It is considered the best solution to the multicarrier networks and proven in [30] by simulations and analysis that it gives better performance than using SHO and HHO separately. It is also so-called Site Selection Diversity Transmission (SSDT). The idea of SSDT is that depending on the channel quality indicator it selects and sends each DL symbol. As shown in [30], the researchers use the SSDT OFDM-based broadband networks with zero-adding to cope the obstacles facing HHO and SHO. For an instance as proven in [30], SSDT has the lower probability of outage comparing to either hard handover or soft handover. Therefore, it is expected that it will be broadly used in a high-speed multimedia services. 38

50 2- Combined SHO and Partial reuse: it is integration between soft handover and partial reuse in the downlink direction of OFDMA system to mitigate the inter-cell interference effect. The target of such mechanism is that the increment of the average throughput specifically at the cell edge while sustaining the data rate fairness among system UEs. This technique is also used to decrease overhead of the SHO. The idea of this system is electing the better signal quality between the SHO system and Partial reuse system for UEs at the cell boundary. 3- Multicarrier Handover Mechanism: this technique can provide an increment in the cell capacity and data rate service. In this system, the UEs can to keep its connection with the source enb while performing the handover with the target enb concurrently which means fast and seamless handover. Figure.3.5 states the multicarrier handover scheme. As shown from the figure, the UEs move from the baste station 1(in LTE enb1) to the base station 2. The carrier 1 is used to keep the connection with the serving BS while the other carrier (Carrier 2) searching for the best target BSs depending on the active target BSs list. At the hysteresis point, the UE performs the handover operation using carrier 2, then disconnecting from BS1 that is made using carrier 1 [29]. The following figure (3.5) illustrates this scheme. Figure 3.5 Multicarrier Handover [29] 39

51 4- Fractional Soft Handover Mechanism (FSHO): this technique divides the services as VoIP and non-voip. This classification of services helps to treat the traffic separately in which the VoIP services use soft handover while the rest of the supported services are utilized of hard handover. It is proven in the simulation in [31] that this scheme is better than SHO in terms of probability of outage and overhead which are both lower. The backward compatibility of this system with LTE gives it a chance to be the preferred option among other HO schemes in order to provide mobility enhancement in LTE-A system Future Trends of Handover One of the main trends in the modern system such as LTE-A is that the fast and seamless handover procedure. It relies on the applied services for instant real-time services (RTS) such as video streaming where there is a need to high data rate and broader bandwidth. This results in reducing the connection for the RTS during the HO process while the users move from serving enb to the target enb. However, this is not the case in the non-rts such as internet browsing in which the need for high data rate and wider bandwidth is unnecessary. The user has not observed any effects during the handover operation [29]. The next factor that could be considered as a future trend of handover is that the backward compatibility and supporting legacy systems such as GSM, UMTS and EDGE. That is; LTE-A and its UE is compatible with the legacy system of telecommunications; thus, its handover techniques have to support the previous communication systems Handover Phases in LTE-A In general, the handover procedure can be divided into four phases as the following: - Initiation phase. - Preparation phase. - Execution phase. - Completion phase. It should be mentioned that some telecommunication resources have divided handover into three phases only, combining Initiation and Preparation in one phase and re-called it Preparation phase. Because LTE-A has two interfaces X2 interfaces and S1 interface, the handover could be 40

52 classified based on these interfaces. While X2-based handover obviously happens between enbs only when there is no need to change the serving MME as a consequence for handover operation, S1-based handover takes place when the MME is changed because of the handover. It further affects the network since it reaches MME node, and it takes more time than the X2-based HO. The main different between those two sorts is the network signaling that happened between source and target enbs and in some cases core network (CN). However, the signaling over the radio link has no change in which same RRC procedure are conducted, and the UE behavior is unchanged [33] 1- Initiation phase: in this phase, the source enb chooses from the neighboring competitive enbs as a target where the UE will switch to. In addition, the serving enb decides when the UE has to be moved to the chosen target enb [33].Figure.3.6 explains the initiation phase. Figure 3.6 X2 Initiation Phase [34] 2- Preparation phase: this is the phase where not only the measurement reports are important as input to the handover decision to be taken, but also the MME could provide another important input to the handover decision. This input is a handover specific list of competitive target enbs used by the serving enb to filter the target enbs [33].The handover decision is important to specify whether the handover is X2 type or S1 handover. X2-based handover in this phase has the following procedure as shown in Figure

53 Figure 3.7 X2 based Handover Preparation Phases [33] Source enb is responsible for initiating the handover request through X2 interface, so it sends a request to the target enb asking for the permission to hand its user and prepare the HO operation. Generally, in the handover preparation phase, the serving enb informs the target enb about all the inter-node RRC-information related to the served UE. This information involves settings of RRC that are already being applied, RRM specific information of the UE, and the information about the connected UE s capability of radio access. These details are required to configure the target enb to be capable of serving the UE during the handover and after completing the HO operation. To recover the probability of handover failure, the source enb includes other important information via inter-node RRC information called re-establishment information used in reestablishing the connection. Via X2 interface, target enb acknowledges the handover request of the source enb directly using Handover Request Acknowledgment message. 42

54 In S1-based handover, on the other hand, the source enb sends the initiated message (Handover Request Message) to the MME through S1 interface informing about the need to trigger the handover preparation with the target cell. Handover Request Message has valuable information such as priority and QoS that it is important to configure the target enb to be ready to serve the transferring UE. Figure 3.8 S1 based Handover Preparation Phases [33] 43

55 While the target enb gets the Handover Request Message, it can accept or refuse the request of handover depending on the feedback in the Handover Request Message. In another words, target enb performs the admission control relying on the available radio resources, enb configuration and information in Handover Request Message. If there is a one affordable enhanced radio access bearer (E-RAB) at the target enb, this enb prepares the required resources to serve the new transferred UE and acknowledges the source enb by sending HO Request ACK message. ACK message is sent to the MME. It contains the required configuration changes that have to be aware by the UE while moving to the target enb. Finally, MME sends back an explicit message called Handover Command message to the source enb containing the RRC connection configuration through S1 interface. S1 based handover is shown in Figure Execution Phase It is a phase in which the target enb commands the source enb to start the handover by sending the RCC reconfiguration message. The source enb forwards without updating the reconfiguration message received from the target enb to the UE. The content of the reconfiguration message is the mobility control info (Handover Command) which is used to order the mobile to reset the current MAC and RRC sessions and to signal with new enb. The Handover Command includes information such as used frequency, target cell downlink and uplink bandwidths, and the target enb physical ID. It also has new C-RNTI that is utilized at a target enb to define the UE and provide the required information to access the common channels (RACH). Moreover; the security information is included in the mobility control info message. If there is data being under transmitting while the handover occurs, the source enb forwards the data to the target enb to prevent data loss. In the X1 based handover, the data is sent through the GTP tunnel directly from the source to target enbs. Unlike X1 based handover, the S1 handover avoids data loss during the handover operation by sending the data via indirect route through S-GW. It should be notice that there is a time limit for handover operation that is set using a timer. When the timer is expired while the handover is still occurring, the UE announces that the handover is unsuccessful and starts the procedure of reconnection establishment to cope that failure. Any delay in handover can be reduced by preventing UE from reading the target cell system information before accomplishing random access operation. After the handover is completed, all 44

56 required system information could be requested by the transferred UE from the target enb. The handover exaction is shown in Figure.3.9. Figure 3.9 Handover Execution Phase [33] 4- Completion Phase It is the phase that from the UE perspective is finished when it sends the RRC connection reconfiguration complete message. In comparison, from the system point of view is that when the network performs further procedures such as releasing the radio resource of the source enb and transferring the data to the target enb. However, different procedures are taken place in this phase regarding whether X1 based handover or S1 based handover. In X1 based handover, the completion indication message called Path Switch Request message is sent by the target enb to MME. Similarly, in S1 based handover, but the message is called Handover Notify message. Upon MME getting either message, it connects with the serving gateway (S-GW) to arrange data switching from the serving enb to the target cell. 45

57 The source enb releases UE and the used radio resources when it is notified by the target enb in X1 based handover or MME in S1 based handover that the handover is completed. This can be conducted using the UE Context Release message or UE Context Release Command message respectively. As aforementioned before, all handover phases are limited by handover timers. These timers are used to guarantee that the handover operation is conducted properly. For instance, if the source enb does not receive a completion messages from other participating nodes (target enb or MME as explained before), the source enb will force the MME to send the UE Context Release Command by sending the Context Release Request message. Figures 3.10 and 3.11 show the completion phase in X1 based handover and S1 based handover respectively. Figure 3.10 Handover Completion Phase-X1 based Handover [33] 46

58 Figure 3.11 Handover Completion Phase-S1 based Handover [33] Admission Control To satisfy SLA and QoS that agreed with the networks customers in the modern telecommunication networks, admission control (AC) is applied which is one of the fundamental and crucial method. AC is not 3GPP standard in which different providers use various AC algorithms to meet their network and customer needs. Therefore, it is specified by the vendor for each enb in the system to guarantee that the newly admitted traffic will not affect the current applied QoS for the served flows [23]. Different restrictions limit AC decision, for example, the required QoS for both new and admitted bearers, the affordable radio resources, and the type of traffic. Mainly, AC operation accepts or rejects the requested EPS - Evolved Packet System bearers in the system. EPS contains a profile that clarifies the QoS requirements involving several numbers of auto-modified parameters. 47

59 The following details clarify EPS auto-modified downlink parameters: 1- QoS Class Identifier (QCI). It is one of the most important parameters that have different values for other parameters such as packet error rate of layer 2, packet latency of layer 2 and priority of scheduling. These parameters can achieve the required HOL delay target by prioritizing different queues. Resources allocation is based on QCI, for example, if UE uses VoIP and browsing services, the higher priority (VoIP) is allocated resources firstly then the browsing. 3GPP has defined nine QCIs with their characteristics as shown in Table 3.1 QCI# Bite Rate Type Priority L2- Packet Error Rate L2-Packet Delay Example services 1 (GBR) ms Conventional voice 2 (GBR) ms Conventional video 3 (GBR) ms Buffered streaming 4 (GBR) ms Real-time gaming 5 (non-gbr) ms IMS signaling 6 (non-gbr) ms Live streaming 7 (non-gbr) ms Buffered streaming, , 8 (non-gbr) ms browsing, file download, 9 (non-gbr) ms file sharing, etc. Table 3.1 QCI Parameters for EPS Bearer QoS Profile [3] 2- Guaranteed Bit Rate (GBR). It is the parameter that grants a certain bit rate to the bearer that is identified as GBR bearer. In the case of non-guaranteed bit rate bearer sort, another parameter called AMBR (Aggregate Maximum Bit Rate) is assigned. A bearer can be allocated a maximum bit rate (MBR) in certain conditions. 3- Allocation Retention Priority (ARP). It is sixteen integer values starts 1 and ends with 16. APR performs admission control decisions prioritization. There is confusion about the different between ARP and QCI. APR relates to services and bearers allocation while as mentioned before, QCI concerns about resource allocation. An example of ARP is that UE aims to setup VoIP (higher ARP priority) along with browsing service; the enb will reject the browsing request and admit only the VoIP request in order not to be overloaded. 48

60 The Radio Resource Management at enb is responsible for managing and handling different load conditions (i.e., low load, moderate load and excessive load). At the excessive load condition, occurrence of a received packet blocking is highly possible. On the other hand, in the situation of low load there is no possibility for packet blocking since the active UEs are few, in turn, the amount of transmitted data is small and the inter-cell interference level is at its minimum value. Moreover, the minimum QoS requirements for the active users are guaranteed. In LTE network, the system starts utilizing all available physical radio resources as the number of admitted UEs into the system increases. Additionally, in order to satisfy QoS constraints for various users, the operation of layer 2 scheduling for more UEs is increased. At a full load using all available PRBs, there is a possibility that the system may admit more UEs while marinating agreed QoS level of the current served users unchanged. The packet scheduler entity will allocate fewer resources to the best effort bearers if the allocated resources for the users with stricter QoS (i.e., GBR) increase. However, further optimizations may be required as the system is loaded with more users. Due to the fact that there are both RT traffic and Non-RT traffic in one scenario, the switching between L2 scheduling (dynamic scheduling) and L3 scheduling (semipersistent scheduling) is beneficial. It is obvious that more regulation is required while the traffic increases. However, the admission control entity will begin blocking arriving new users traffic although the RRM functionalities (i.e., scheduling) aim to increase cell capability to serve more users with their associated traffic types [24] Packet Scheduling (PS) Radio Resource Management entity at the enb for a multi-carrier advanced LTE network is shown in Figure Essentially, the main two parts of RRM are: carrier component (CC) allocation and Packet Scheduling (PS). Carrier component allocation is that RRM selects and allocates CCs for each UEs while Packet Scheduling (PS) is responsible for assigning radio resources to each user within each CC. The PS decision is made at each transmission time interval TTI (1 ms) or at resource-block-pair (RB of o.5 ms subframe over 180 khz), taking into account the feedback from the users using Channel Quality Indicator CQI. It helps enb to estimates the reachable throughput for each user using the feedback information. Furthermore, enb informs the users about the affordable allocated resources [25]. Even though the load status and the user past throughput are present in the enb, only uplink CQI feedback is useful from the enb point of view to make a decision about resources allocation. 49

61 One the other hand, each UE measures the received signal to interference ratio (SINR) carried on the reference signal sent by the serving base station in the downlink direction. A user is usually moving in the cell coverage; thereby the time-selective fading and multi-path fading natures exist. This results in different calculated SINR values on each subcarrier at each Transmission Time Interval. The measurements, specifically effective SINR values, aid UE to feedback its channel status to the serving enodeb. The values of effective SINR are used by the base station to select modulation and coding scheme (MCS) in the downlink packet direction [35].The later (MCS) is related to the data rate determination, in which the bits that are supported by users are determined based on MCS. Not only finding out the data rate in two contiguous RBs as mentioned before, but also selecting the priority of the users in channel-depending scheduling is supported using the effective SINR. The packets for each user are buffered upon arriving at enodeb, and time stamped by the scheduler. Then, the First-In-First-Out (FIFO) technique is used to handle user s packets. To control the queuing operation and avoid long waiting time for the packet, Head Of Line (HOL) packet latency has been utilized. It is the different between the current time of a certain packet and arrival time of the same packet [35]. The HOL delays are assigned for different traffic types based of the classification of the traffics as Real Time (RT) and Non-Real Time (NRT) services. The threshold point of the HOL is when it exceeds the delay s deadline; at this point the queued packets are deleted. Figure 3.12 RRM Framework in LTE-A [25] 50

62 Packet scheduling can be classified into two types: in the downlink direction which is the most important one since it is related to enb and performed by RRM, and in the uplink direction which is conducted by UE Downlink Packet Scheduling Due to the fact that there is a limited number of available radio resources for each network operators, PS is proposed to meet the goal of maximizing the utilization of these resources, and in the meantime satisfying the agreed level of QoS for connected UEs. The decision of scheduling is made at each transmission time interval; hence, UEs are allocated different amount of radio resources blocks (PRBs) each TTI according to their requested services. The decision not only includes PRBs allocation, but also Modulation and Coding Schemes (MCS) or what socalled link adaptation to be used in the downlink packets communication. PDCCH is used to carry the allocated resources to the users. The link between the enb and users has two main L2 flows: one carries the data (data plane), and one carries the control information (control plane). In the downlink direction, there is interaction between the packet scheduler and Hybrid ARQ entity as shown in Figure Retransmission is managed by Hybrid ARQ manager and scheduled dynamically by the PS in frequency/time domains. While the frequency domain scheduling means that the user is allocated PRBs, the time domain scheduling means that the user is selected to be scheduled at TTI. The scheduler serves all UEs and ensures there is fairness among them by sending either a new flow or awaiting Hybrid ARQ flow to each scheduled UE in one TTI. If the scheduler schedules both flows simultaneously to one user, other users in the system will suffer starvation [24]. Figure 3.3 also shows Link Adaptation mechanism which interacts with scheduling operation to provide the suitable modulation and coding schemes (QPSK, 16QAM or 64 QAM) according to the utilized physical resource blocks for each UE. The CQI receiving from the users and QoS at the enb are used in LA decisions [3]. 51

63 Figure 3.13 Interactions between HARQ, PS and LA [3] The higher channel quality user is the selected user to be scheduled in the frequency domain scheduling schemes (FDPS) since frequency selective fading is achieved. Accordingly, any PRBs with deep fade are avoided by FDPS [24] as shown in Figure Figure 3.14 Frequency DPS Concept [3] There are two methods that have been used to conduct the packet scheduling: per each carrier component (CC) so-called independent scheduling or cross CC scheduling. The cross CC scheduling is more complex than independent scheduling since it based on all other available CCs in the system. This is; the metric of scheduling is calculated differently in each one [25]. 52

64 Independent CC PS. the similarity to the traditional PS in a single carrier system is shown in this method, in which there is no need to consider the transmission characteristics on other CCs. Dividing the instantaneous throughput by the average throughput of the selected user is the used method to calculate the scheduling metric. (3.1) the estimated throughput for user k on i th CC at the j th PRB group is represented by, the average throughput at the past for the same user on the same CC is represented by. The equation is considered to Rel8. For LTE-A, since it could be assigned more than one CC, the same equation is used multiplied by total number (N) of allocated CCs for LTE- A UEs. Cross-CC PS. considering the transmission characteristics of all CCs, PS fulfills better resource allocation than independent packet scheduling. Unlike independent PS, the past user throughput over the all aggregated CCs is taken into account to calculate the scheduling metric. (3.2) Packet Scheduling Algorithms in Downlink Direction The packet scheduling algorithms are various based on RT and NRT services. The common adopted PS algorithms according to [26] and [35] are: First-In-First-Out (FIFO). The user with the highest packet delay HOL at each TTI is given transmission priority. It provides considerable fairness among users who have similar packet s characteristics such packet size and channel status [35]. Round Robin (RR). In this algorithm, transmission time interval is divided equally among users, in which each user is allocated equal time to transmit its packet in a circular order. This algorithm is similar to FIFO that provides a high level of fairness. However, the throughput performance in RR is higher than FIFO. 53

65 Maximum Rate scheduler (Max Rate).Once the highest achievable data rate is reached, the UE is selected to be scheduled. Equation 3.3 expresses Max Rate scheduling algorithm: (3.3) is the reachable data rate for a UE denoted by k at time t based on the received SINR at enb where and SINR are directly proportional. Because this algorithm is designed to schedule users with maximum data rate regarding with their best channel conditions, it provides the best system throughput. Accordingly, the poor fairness is obvious in Maximal Rate PS algorithm since the lower received SINR value the less opportunity for user to be selected for transmission. Moreover, UE s resources starvation could happen since the user is probably never selected for scheduling. Proportional Fair (PF). PF algorithm balances the system performance between the throughput and the fairness among users, giving a trade-off between them. A user is allocated resources if it has the maximum ratio between the instantaneous achievable rate and the transmission rate divided by the average throughput. (3.4), is the average throughput of user k at time slot t. It is calculated by considering the window size as follows: (3. 5) when UE k is selected for transmission, the value equals 1, otherwise the value equal 0. The reason behind a good throughput and fairness performance is that PF algorithm performs incorporation for the feasible data rate with an average throughput [35]. Frame Level Scheduler (FLS). The FLS is a combined algorithm that has two levels of scheduling; upper and lower level, thereby, separate algorithm for each one. Upper level is less complexity in allocation of resources since it depends on the theory of discrete time linear control. The task of the upper level is that satisfying delay constraint by calculating the amount of data that each real-time source should send within a single 54

66 frame. Equation 3.16 shows the calculation of the aforementioned amount of data at upper level of FLS. The lower level, on the other hand, is more complexity since it uses PF algorithm to allocate resources to the UE [36]. (3.6) denotes the data amount that is sent within the frame n for flow. is the filtered signal by which is a time-invariant linear filter with pulse response. According to the results in [11] cited in [36], the FLS provides restricted delays and lower PLR values for a video traffic,thereby, the performance of FLS ensures the best quality of the video service for the scheduled users. Modified-Largest Weighted Delay First (MLWDF). This algorithm is introduced to support Real Time services. The metric of scheduling is shown : (3.7) where; (3.8) HOL packet delay of user at time is denoted by, is the delay s deadline that the maximum HOL probability to exceed it is. The benefit of MLWDF is that using PF properties along with the HOL packet delay consideration introducing a better throughput, fairness and Packet Loss ratio than PF algorithm. Exponential/Proportional Fair (EXP/PF). The aim of proposing this algorithm is to be used for multimedia services for both Real Time and Non-Real Time services simultaneously. The metric for RT and NRT is shown below: (3.9) where, (3.10) (3.11) 55

67 denotes the average number of packets queued in the buffer at time t, k and in equation (3.11) are predefined values, is the highest HOL packet delays for the packet awaiting for RT services, and is the maximum delay of RT service users. Although EXP/PF is adopted for both real time and non-real time users, it precedes RT users over the NRT if the RT users packets reach the transmission threshold[35]. Logarithmic (LOG) Rule scheduler. The LOG Rule allocates the resources to the scheduled UEs increasing current throughput by supposing the channel status and traffic arrival are realised. The main propose that LOG Rule is designed to provide QoS balancing in term of robustness and mean delay. According to [13] cited in [36], the simulation results prove that LOG Rule is a superior algorithm that has the best packet delay decreasing. Although this algorithm is an experimented practically as a good solution, it is not proven as an optimal method for mean-delay achievement [36]. LOG Rule and EXP Rule algorithms are a type of opportunistic scheduling, in which they exploit the desirable channel to schedule the active users, for example, the users with the highest rate. Exponential (EXP) Rule. Similarly to LOG Rule, the goal of proposing this algorithm was basically to satisfy QoS requirements in the wireless network. The balance between the throughput and mean-delay is conducted by maintaining minimum value of delay and at the mean time a considerable value of throughput, thereby, mean-delay is obtained. It works on the concept of wireless channel sharing among arrival users and queues their data as a random stream to be prepared for transmission. This scheduler enables users accessing the services during each time interval. Optimal throughput is determined according to [12] cited in [36]. The difference with LOG rule is that there is no previous estimation of traffic arrival and channel statistics. However, the EXP Rule explicitly utilizes received channel statistics information and guarantees achieving stable queues. Two exponential rules have been introduced in this scheme: EXP Waiting time (EXP-W) and EXP Queuing (EXP-Q). The algorithm selects one rule at a time for scheduling users with different fix positive parameters (i.e., and ) as shown in equations (3.12) and (3.13). 56

68 (3.12) (3.13) where, and (3.14) is the total number of queued users who are selected for transmission at time Uplink Packet Scheduling In the UL scheduling, the user is aware of packets scheduling, and it has to buffer the arrived flows. The finite size of the users buffers degrades the performance of scheduling operation at the UEs since the base station in LTE is unaware of the size of UEs buffer. Not only the buffering limitation at the UEs, but also the power limitation in uplink direction emerges another constraint for the uplink scheduling. Obviously, enb power in the downlink is much higher than the UEs power. On the other hand, resource allocation restriction exists since single carrier modulation has been used with uplink scheduling. In consequence, solely adjacent PRBs can be assigned to each UE. CSI report is the feedback information that is used to choose the modulation and coding scheme. Relying on the Sounding Reference Signal (SRS) sent by the UE, the CSI is determined. The integration in the uplink direction between RRM, LA and PS is shown in the Figure Link Adaptation compromises adaptive modulation and coding, Outer Loop Link Adaptation (OLLA) and Power Control (PC). PS has scheduling request, buffer status report and Adaptive Transmission Bandwidth (ATB). There is also a relation as seen in Figure.3.15 between LA and PS, in which PC and AMC of LA interact with PS. That is, on the channel state, the packet scheduler receives required information related to the transmission bandwidth from AMC for a certain user. Uplink PC main purpose is that maintaining SINR to a certain level called SINR threshold according to the agreed QoS while limiting the ISI [3]. 57

69 Figure 3.15 Uplink RRM Functionalities inter-work with LA and PS [3] Power Control (PC) The Orthogonal Frequency Division Multiple Access OFDMA and the Single Carrier Frequency Domain Multiple Access are the radio technologies used in LTE and LTE-A. The reason behind adopting OFDMA and SC-FDMA in the modern networks is that to elevate the effect of intracell interference that is the interference between users within a single cell. The orthogonality is used to avoid users having the same peak at a certain point of time instead only one user could be served at that peak. However, another interference causing by adjacent cells, which is wellknown as inter-cell interference cannot be negligible, introducing a real challenge since orthogonal modulations is not designed to solve such interference. It requires other mechanisms to solve it that the power control is one of them. Since the PC limits the cell boundary, it can eliminate the impact of inter-cell interference. The control of the transmitted power can be performed at UE in the uplink direction or the base station (enodeb) in the downlink direction [32]. 58

70 The UE transmits the power in the uplink direction based on the equation (3.15). (3.15) Where and are the UE s maximum allowed transmit power based on the UE power classification and the number of allocated physical radio resources on the PUSCH respectively. UE measures the downlink path loss which is denoted by, denotes closed-loop power control correction that enb transmits (details can be found in of [37] ). are parameters of PC. The can be computed as follows [32]: (3.16) where is the target of open loop SNR (details can be found in of [37]), and are the PRB noise power and the number of PRBs required to fulfill the target SNR with full power, respectively. UE initiates a transmission power based on and the calculation of the path loss performed by the UE. Since enb signals the value of to UE who already has completed setting the initial power, it does not contribute in the initiation operation. Hence, by ignoring the value of as well as, the equation (3.15) can be written as following: [dbm] (3.17) It represents the calculation of the initial transmitted UE power. The number of scheduled PRBs is denoted by in which the UE allocates power based on PRB. That is; the amount of power for each PRB is equal. To calculate the UE s Power Spectral Density (PSD) to each PRB, the value of is neglected and the equation (3.17) is changed to be as the following [32]: [dbm/prb] (3.18) The power control has two types depends on the value of in Equation (3.18). If the value of is between 0 and 1, the power control mode is called fractional power control. When the value of is 1, the power control mode is called conventional power-control. Other types of power control have to be mentioned here which they are: open loop power control and close loop power control. Detailed descriptions of the power control modes can be found in [37]. 59

71 Balancing of Carrier Load LTE and LTE-A can coexist together in the same network due to the backward compatibility of LTE-A. As mentioned before the maximum prescribed bandwidth of LTE is 20MHz. To achieve LTE-A bandwidth that is 100 MHz, 5 CCs of Rel8 have to be aggregated. In LTE, UEs are allocated a single CC while in LTE-A UEs are assigned multiple CCs based on their channel conditions. The flowchart in Figure.3.16 explains the LTE-A s enb classification operation whether the case of LTE UE or LTE-A UE to make a decision of radio resources allocation on CC(s). To balance the load on all available CCs, UEs are equally scheduled on all CCs using a smart load balancing mechanism. This could guarantee an exploitation of all affordable resources on CCs. The following sections discuss the load balancing methods. Figure 3.16 enb Classification for LTE Rel 8 and LTE-A Arrival UEs [25] Carrier Load Balancing As aforementioned, Rel 10 has a backward compatibility with Rel8. For a user in LTE, the system allocates one CC while the system could allocate number of CCs based on the QoS and user s feedback reports for a user in LTE-A. The CCs balanced distribution among the served users can be performed by deploying balancing methods as the following [27]: 60

72 Round Robin (RR) Balancing Method: in this method, the user is allocated 1 CC. The newly arrived user is assigned the least exploited carrier by other users. The aim of that is to divide UEs equally on all available CCs. An issue of this method is that the CCs load could have a minor difference due to the probability of uneven number of users or due to the fact that the users could free up the allocated CC at random. Mobile Hashing (MH) Balancing Method: also well-know independent carrier channel assignment. Similarly to RR, a user is allocated 1 CC. However, MH differentiates from RR by using hash algorithm calculations of the terminals. The purpose of considering the output of hash values is that providing a long-term CCs load balancing. In order to achieve that, it requires uniformly distributing of the values of hash outputs among a limited set that CC indices are mapped directly on it Interference Management The interference is one of the big challenges encounter LTE and LTE-A especially with HetNets, in which the small cells (Pico, Femto, RRH or relay node) use the same carrier frequency of the Macro cells. The higher power enb (Macro cell) is overlaid with small lower power cells that are used with less care or uncoordinated manner. The technology that has been implemented to relieve the impact of the interference between adjacent cells is the Inter-cell Interference Coordination (ICIC). ICIC is the operation in which the interference could be mitigated if high transmission power on PRBs is avoided. That is; the users on the cell edge can be served in the neighboring cells. There are two classes of ICIC schemes based on the way it deals with the interference. The first class is reactive ICIC that is responsible for monitoring the system. That is; if it observes a high level of the interference, the suitable procedures will be implemented. Examples of the reactive ICIC are packet scheduling and power control for the purpose of interference reduction to an appropriate status. The other class is proactive that is responsible for avoiding the interference before the high level is detected. That could be performed through the enbs coordination in which neighboring enbs receives feedback from the enb informing about the future plans of scheduling its users. These reports can be considered to avoid low value of Signal- to- Interference Ratio (SIR) that could occur [3]. There is a relation between proactive ICIC and Relative Narrowband Transmit Power (RNTP). RNTP is a PRB s peak downlink transmission power. Neighboring enb receives RNTP through X2 interface, then these RNTP 61

73 values can be utilized by the neighboring enb to make a decision for scheduling its UEs. Especially, the UEs who are more likely existed at the cell edge have a high probably to suffer from neighboring cells interference (LTE case) or small cells and neighboring cells (LTE-A case).thus, RNTP facilitates the proactive ICIC in the downlink direction. Various parameters are considered to perform the proactive and reactive ICIC in the uplink direction. The High Interference Indicator (HII) and Overload Indicator (OI) are used to support proactive and reactive ICIC schemes respectively. HII carries a serving enb message to its adjacent cells over X2 interface indicating which PRBs will be exploited for scheduling boundary cell UEs such as the higher interference expectation of PRBs from the neighboring cells perspective. Hence, the adjacent cells allocate those PRBs to the lower interference UEs. This is the reason that why HII is seen as a technique of proactive ICIC. As mentioned before, OI is related to reactive ICIC scheme and its task essentially is to carry reports of the measured uplink interference on reporting enb s PRBs at three levels (low, average and high). The adjacent enbs deal with these measurements once they are received by adjusting the scheduling behavior to the extent that enhances the SIR of the OI releasing enb [11] Summary RRM is the entity that is mainly responsible for the following: Handover, Admission Control, Packet Scheduling, Power Control and Interference Management. Thus, it plays a vital role in the most recent LTE and LTE-A networks where the most important functions in managing a mobile network is performed at enb using RRM. While the LTE-A is introduced through HetNets, the importance of RRM has increased due to the increment of the issues related to interference, handover and scheduling. Since the simulation and practical outcomes in the next chapter mainly concern with Handover and Packet Scheduling and its algorithms, the focus in this chapter is on these RRM functionalities. 62

74 References [1] H. Holma and A. Toskala, Wcdma for Umts vol. 4: Citeseer, [2] A. Ghosh and R. Ratasuk, Essentials of lte and lte-a: Cambridge University Press, [3] H. Holma and A. Toskala, LTE for UMTS: OFDMA and SC-FDMA based radio access: John Wiley & Sons Inc, [4] K. Sandrasegaran. (2012, 5 Aug 2012). 4G and Long Term Evolution (LTE). Available: < > [5] H. G. Myung, "Introduction to single carrier FDMA," [6] J. Hågstrand, P. Karlsson, and O. Silverplats, "Proof of concept implementation of UMTS long term evolution," [7] M. Rumney, "3GPP LTE: Introducing Single-Carrier FDMA," Agilent Technologies White Paper, [8] H. Holma and A. Toskala, LTE Advanced: 3GPP Solution for IMT-Advanced: Wiley, [9] A. Lucent. (2009, 20 Aug 2012). The LTE Network Architecture A comprehensive tutorial. Available: [10] S. Sesia, I. Toufik, and M. Baker, "LTE The UMTS Long Term Evolution," From Theory to Practice, published in, vol. 66, [11] E. Dahlman, S. Parkvall, and J. Sköld, 4G LTE/LTE-Advanced for mobile broadband: Academic Press, [12] J. Löfgren. (2009, 3 Sep 2012). LTE Radio Interface ArchitectureAvailable: < > [13] H. G. Myung, "Technical overview of 3GPP LTE," Polytechnic University of New York, [14] M. Sauter, FROM GSM TO LTE AN INTRODUCTION TO MOBILE NETWORK AND MOBILE BROADBAND. United Kingdom: John Wiley & Sons Ltd, [15] C. Cox, AN INTRODUCTION TO LTE, LTE-ADVANCED, SAE AND 4G MOBILE COMMUNICATION. United Kingdom: John Wiley & Sons Ltd, [16] E. Dahlman, 3G evolution: HSPA and LTE for mobile broadband: Academic Press, [17] D. Kimura and H. Seki, "Inter-Cell Interference Coordination (ICIC) Technology," FUJITSU Sci. Tech. J, vol. 48, pp , [18] B. Y. B. Cho. (2011, 23 Aug 2012). LTE DL Transmission. Available: < > [19] 3GPP. (2011, 5 Sep 2012). Technical specification Group Services and System Aspects; Multimedia Broadcast/Multicast Service (MBMS); Architecture and functional description version (Release 9). Available: < > [20] 3GPP. (2011, 15 Aug 2012). Technical Specification Group Radio Access Network; Feasibility study for Further Advancements for E-UTRA (LTE-Advanced) version (Release 10). Available: < > [21] I. F. Akyildiz, D. M. Gutierrez-Estevez, and E. C. Reyes, "The evolution to 4G cellular systems: LTE-Advanced," Physical Communication, vol. 3, pp ,

75 [22] 3GPP. (2010, 10 Sep 2012). Technical Specification Group Radio Access Network; Evolved Universal Terrestrial Radio Access (E-UTRA); Further advancements for E- UTRA Physical layer aspects version (release 9). Available: < > [23] E. Z. Tragos, G. Tsiropoulos, G. T. Karetsos, and S. A. Kyriazakos, "Admission control for QoS support in heterogeneous 4G wireless networks," Network, IEEE, vol. 22, pp , [24] K. I. Pedersen, T. E. Kolding, F. Frederiksen, I. Z. Kovács, D. Laselva, and P. E. Mogensen, "An overview of downlink radio resource management for UTRAN long-term evolution," Communications Magazine, IEEE, vol. 47, pp , [25] Y. Wang, K. I. Pedersen, T. B. Sørensen, and P. E. Mogensen, "Carrier load balancing and packet scheduling for multi-carrier systems," Wireless Communications, IEEE Transactions on, vol. 9, pp , [26] A. Hecker, M. Garcia-Lozano, and J. F. Monserrat, "Resource Management in 4G Networks," Pervasive Mobile and Ambient Wireless Communications: COST Action 2100, p. 461, [27] Y. Wang, K. I. Pedersen, P. E. Mogensen, and T. Sorensen, "Carrier load balancing methods with bursty traffic for LTE-Advanced systems," 2009, pp [28] Y. Lee, J. Zhao, and H. Qu, "A combined handover scheme for LTE-advanced system," 2012, pp [29] I. Shayea, M. Ismail, and R. Nordin, "Advanced handover techniques in LTE-Advanced system," 2012, pp [30] H. Lee, H. Son, and S. Lee, "OFDM-Based Semi-Soft Handover for High Data Rate Services," in Personal, Indoor and Mobile Radio Communications, PIMRC IEEE 18th International Symposium on, 2007, pp [31] J. Chang, Y. Li, S. Feng, H. Wang, C. Sun, and P. Zhang, "A fractional soft handover scheme for 3GPP LTE-advanced system," in Communications Workshops, ICC Workshops IEEE International Conference on, 2009, pp [32] B. Muhammad, "Closed loop power control for LTE uplink," Blekinge Institute of Technology School of Engineering, [33] June Yi, S.C., YoungDae Lee, SungJun Park, SungHoon Jung, Radio Protocols for LTE and LTE-Advanced, [34] Penttinen, Jyrki, LTE/SAE Deployment Handbook, [35] Nguyen, Sinh Chuong, Packet Scheduling for LTE-Advanced, Master thesis [36] Shaoo, Biswapratapsingh, Performance comparison of packet scheduling algorithms for video traffic in LTE cellular network, PInternational Journal of Mobile Network Communications & Telematics ( IJMNCT) Vol. 3, No.3, June [37] AL-Jaradat, Huthaifa, Radio Resource Management in LTE and LTE-A, Master thesis

76 4. Chapter 4: LTE-Sim Heterogeneous Network Deployment 4.1. Introduction In the Long Term Evolution so-called LTE, the requirements for larger coverage area, more capacity, and high data rate and low latency have led to search for cost-effective solutions to meet these demands. Hence, the development in the telecommunication networks has adopted different directions to enhance the LTE system taking into account the International Mobile Telecommunications (IMT-2000) standards that have to be satisfied [1]. Network-based technologies such as Multiple Input and Multiple Output MIMO/ advanced MIMO and Transmission/Reception Coordinated Multi-Point CoMP are LTE enhancements that introduce LTE Advance (LTE-A). Other fewer cost enhancements based on air interfaces are proposed, such as improving spectral efficiency involving using Heterogeneous networks (HetNets). HetNets are small and lower power cells within the main Macro cells with different access technologies to close up the network to the end users and increase their expectation [16].According to [2], there are two main practical HetNets classes: Macro with Femto and Macro with Pico. Femto and Pico are the small and lower power cells. To save the cost, operators use the same carrier frequency in the large and small cells which, on the other hand, proposes interference challenges. Figure.4.1 gives the main concept of HetNets. To clarify, user in LTE is well-known as a UE. Figure 4.1 an Example of HetNets 65

77 In LTE and LTE-A, the element that is responsible for Radio Resources Management (RRM) is enhanced Node Base station (so-called enb). The enb does all required management including Packet Scheduling (PS) which is the focus in the paper. PS can guarantee the agreed quality of service demands (QoS) because it is responsible for the best and effective utilizing of the affordable radio resources and in charge of data packets transmission of the users[3]. 3rd Generation Partnership Project (3GPP) has left the scheduling algorithms to be a vendor specific according to user s requirements and network capability. Therefore, various PS algorithms have been proposed depending on the traffic sorts and provided services. PF, MLWDF and EXP/PF algorithms [4][5][6] are used in this paper to study and compare between the system behaviours in HetNets (single Macro with 2 Pico cells) using these three types of algorithms. Scheduling algorithms ensure that QoS requirements have been met. This can be conducted by prioritizing each link between the enb and the users, the higher priority connection the first handled in the enb Downlink System Model of LTE The basic element in the downlink direction of the LTE networks is called Resource Block (RB).Each UE is allocated certain number of resource blocks according to its status, the traffic type and QoS requirements. It could define the RB in both frequency domain and time domain. In the time domain, it comprises single (0.5 ms) time slot involving 7 symbols of OFDMA (orthogonal frequency division multiple access). In the frequency domain, on the other hand, it consists of twelve 15 khz contiguous subcarriers resulting in 180 khz as a total RB bandwidth [7]. As aforementioned before, the enb is responsible for PS and other RRM mechanisms. The bandwidth that is used in this study is 10 MHz considering the inter-cell interference exists. The period that enb performs new packet scheduling operation is the Transmission Time Interval (TTI). TTI is 1 ms that mean the users are granted 2 contiguous radio resource blocks (2RBs). The scheduling decision in the serving enb is made based on the uplink direction reports come from the UEs at each transmission time interval. The reports comprise the channel conditions on each RB, such as signal to noise ratio (SNR). The serving enb uses the SNR value involved in the reports to specify the DL data rate for each served UE in each TTI. For example, how many bits per 2 contiguous RBs [8]. 66

78 The data rate for user i at j sub-carrier on RB and at t time can be determined by using equation (4.1) as proposed in [9]. (4.1) A = B = C = D = rgg The number of bits per symbol is A. The number of symbols per slot is B. While C represents how many slots per TTI, D clarifies how many sub-carriers per RB. Table 4.1 summarizes the mapping between SNR values and their associated data rates. Minimum SNR Modulation and Data Rate Level (db) coding (Kbps) 1.7 QPSK (1/2) QPSK (2/3) QPSK (3/4) QAM (1/2) QAM (2/3) QAM (3/4) QAM (2/3) QAM (3/4) 756 Table 4.1 Mapping between instantaneous downlink SNR and data rate Upon the packets reach the enb, they are buffered in enb in a specific container allocated for each active UE. Moreover, the buffered packets are assigned a time stamp to ensure that they will be scheduled or dropped before the scheduling time interval is expired, and then using First- In-First-Out (FIFO) method they are transmitted to the users in the downlink direction. To explain the scheduling operation, PS manager (is a part of enb functionalities) at each TTI priorities and classifies the arriving users packets according to preconfigured scheduling algorithm. 67

79 Scheduling decision is made based on different scheduling criteria that have been used in various algorithms. For example channel condition, service type, Head-of-Line (HOL) packet delay, buffer status, and so on so forth. One or more RBs could be granted to the selected user for transmission with the highest priority. Figure.4.2 shows the packet scheduler in the downlink direction at enb. Figure 4.2 Downlink Packet Scheduler of the 3GPP LTE System [10] 4.3. Packet Scheduling Algorithms The efficient radio resource utilization and ensuring fairness among connected users, as well as satisfying QoS requirements, are the main purposes of using PS algorithms [11].The PS algorithms that have been used in this study are : Proportional Fair (PF) algorithm, Maximum- Largest Weighted Delay First (MLWDF or ML) and the Exponential/Proportional Fair (EXP/PF or EXP) algorithm. It should be noted that these algorithms are used Proportional Fair (PF) Algorithm For non-real time traffic, the PF was proposed which is used in a Code Division Multiple Access- High Data Rate (CDMA-HDR) system in order to support Non-Real Time (NRT) traffic. In this algorithm, the trade-off between fairness among users and the total system throughput is presented. 68

80 This is, before allocating RBs, it considers the conditions of the channel and the past data rate. Any scheduled user in PF algorithm is assigned radio resources if it maximizes the metric k that calculated as the ratio of reachable data rate of user i at time t and average data rate of the same user at the same time interval t: where; (4.2) (4.3) is the window size used to update the past data rates values in which the PF algorithm maximizes the fairness and throughput for any scheduled user. Unless user i is selected for transmission at, = Maximum Largest Weighted Delay First (MLWDF) Algorithm If the traffic is a Real Time (RT), the MLWDF is introduced which is used in CDMA-HDR system in order to support RT data users [11].It is more complex algorithms compare with PF and is used in different QoS user s requirements. This is because it takes into account variations of the channel when assigning RBs. Moreover, if a video traffic scenario, it takes into consideration time delay. Any user in MLWDF is granted RBs if it maximizes the equation below: (4.4) where; (4.5) where is a difference in time between current and arrival times of the packet that known as the Head Of Line (HOL) packet delay of user i at time t. Similarly to PF equation, while the achievable data rate of user i at time t is, the average data rate of the same user at the same time interval t is. and are the delay threshold for a packet of user i and the maximum HOL packet delay probability of user i respectively. The later is considered to exceed the delay threshold of user i. 69

81 Exponential/Proportional Fair (EXP/PF) Algorithm Since PF is not designed for multimedia applications (only for NRT traffic), an enhanced PF called EXP/PF algorithm was proposed in the Adaptive Modulation and Coding and Time Division Multiplexing (AMC/TDM) systems. The EXP/PF algorithm is designed for NRT service or RT service (different sorts of services). The metric is used for both RT nad Non-RT in which RBs are assigned to users based on. (4.6) where, (4.7) (4.8) where the average number of packets at the buffer of the enb at time t is represented by, k and in equation (8) are constants, is explained in MLWDF, is the HOL packets delay of RT service and is the maximum delay of RT service users. The EXP/PF differentiates between RT and NRT by prioritizing RT traffic users over the NRT traffic users if their HOL values are reaching the delay threshold Simulation.1- Single Macro Cell with two Pico Cells This simulation has been performed to compare between a telecommunication system that involves only Macro cell and same telecommunication environment with adding two small low power cells called Pico cells. 70

82 Simulation.1 Environment LTE-Sim simulator is used to do the entire analysis and study [12]. The most recent version of LTE-Sim (version 5) has not involved yet any code regarding the HetNets type (Macro with Pico cells). The developed code used in this study could be considered as an enhancement of the released LTE-Sim versions. However, LTE-Sim has a detailed code (or what authors are named it: scenario) which can be used to simulate and examine HetNets type (Macro with Femto). This simulation is based on a scenario of a single Macro cell with 2 small Pico cells that are reduced their powers. More Picos can be added to the system, and enhanced system behaviour will be presented (details are discussed in simulation 3). However, according to [2], while the number of Pico cells is increased, more inter-cell interference is experienced since the same carrier frequency is used in each cell (Macro and Picos). Figure.4.3 shows the entire system that is deployed: Macro cell of 1 km and 2 Pico cells of 0.1 km located on the Macro edge. This design is chosen to emulate a real system aimed to cover larger area and more users, especially the users at the cell edge where they suffer from lack of connectivity with Macro cell. The inter-cell interference is modeled. Video and VoIP traffic are used to represent user s data. Each user has 50 % Video traffic and 50% VoIP flows. Handover is activated. Each cell starts a certain number of users. Non-uniform user distribution within the cells is deployed, and 3km/h constant speed is utilized as the user speed mobility. In addition, the 3GPP urban Macro cell propagation loss model has been implemented including path-loss, penetration loss, multi-path loss and shadow fading which are summarized below [13]: Pathloss:, d refers the distance between the enb and the user in kilometers. Penetration loss: 10 db Multipath loss: using one of the well-known methods called Jakes model Shadow fading loss (recently it could be used as a gain in LTE-A): log-normal distribution - Mean value of 0 db. - Standard deviation of 10 db. 71

83 Figure 4.3 Applied HetNets (Macro with 2 Picos) Packets throughput (see equation 4.9), Packet Loss Ratio (PLR) as viewed in equation (4.10), packet delay (latency) and fairness index (equation 4.11) are the concepts used in the aforementioned algorithms to evaluate the system performance. Jain s method is applied to implement fairness among users [14]. According to [1], fairness should reach the value of 1 to be considered as a fair algorithm that sharing the resources suitably among users. It can be calculated as the value of one minus the value of the difference between the maximum and minimum size of transmitted packets of the most and least scheduled users. Equation (4.11) calculates the fairness value. (4.9) (4.10) (4.11) 72

84 Obviously, while is the size of transmitted packets, is the size discarded or lost packets during the connection. is the summation of all arrived packets that are buffered into serving enb [1]. The aforementioned total size of transmitted packets of the best served UE and the worse served UE are represented in equation (11) as and.table 4.2 shows the entire system simulation parameters [1]. Parameters Simulation time 30 s Flow duration 20 s Slot duration 0.5 ms TTI 1 ms Number of OFDM symbols/slot 7 Macro cell radius 1 km Macro enb Power 49 dbm Pico cell radius 0.1 km Pico enb Power 30 dbm User speed 3 km/h VoIP bit rate 8.4 kbps Video bit rate 242 kbps Frame structure type FDD Bandwidth 10 MHz Number of RBs 50 Number of subcarriers 600 Number of subcarriers/rb 12 Subcarrier spacing 15 KHz Table 4.2 LTE System Simulation Parameters 73

85 In order to get better results and confirm the outcomes, five simulations have been conducted for each algorithm (PF, MLWDF and EXP) in each point of users (10, 20, 30, 40, 50, 60, 70 and 80). This yields 120 simulations outcomes. The average values have been taken to draw the simulation graphs at each point of users Simulation.1 Results The system is judged base on throughput, Packet Loss Ratio (PLR), delay and fairness Throughput The average overall system throughput is shown in Figure.4.4. Comparing the throughput for single Macro cell for the same simulation parameters as viewed in Figure.4.5, the pico cells in the scenario Macro with 2 Picos boost the throughput by adding gain that shown as an overall system throughput increment for the same number of users. For instance, at 40 users using MLWDF, the throughput is 25 Mbps for the scenario with 2 Picos while the Macro scenario is only 9.3 Mbps. This is almost a double value. Further points show duple and triple throughput values in the scenario of 2 Picos. However, the gain will reach a saturation level where no more gain could be obtained due to the fact of limited radio resources availability while more users are added to the system. Although MLWDF and EXP have almost similar behaviour in both scenarios, a higher throughput is acquired in the 2 Pico case using both algorithms. It could note that PF algorithm behaves better than the scenario of single Macro cell as seen Figure.4.5. PF is developed for NRT traffic, but the simulation is for Video flows (RT traffic); hence, the other simulated algorithms outperform PF. Figure 4.4 Average System Throughput (Macro with 2 Picos) 74

86 Figure 4.5 Average System Throughput (single Macro cell) Packet Loss Ratio (PLR) PLR shown in the Figure.4.6 according to [15] is the packet loss ratio for a single Macro cell. While the system is charged with more than 20 users, the PLR is increased for all experienced algorithms taking into consideration that the PF is the worst case with the video traffic. Adding two Picos to the previous system to create Macro with 2 Picos scenario enhances the PLR while maintaining similar system behavior for all algorithms. Approximately, the PLR in Macro with 2 Picos case is reduced to be a quarter of PLR value of single Macro cell scenario. For example, at 70 users, MLWDF has 0.1 PLR value while for the same number of users MLWDF has 0.5 PLR value in the single Macro scenario. PF algorithm is the worst case in both simulated cases comparing with the other scheduling schemes. Figure.4.7 illustrates PLR for Macro with 2 Picos. Figure 4.6 PLR of Video Flows (single Macro cell) [15] 75

87 Figure 4.7 PLR of Video Flows (Macro with 2 Picos) Delay According to [15] and as illustrated in Figure.4.8, the delay in single Macro cell scenario is close to be constant for PF, MLWDF and EXP/PF with value less than 5 ms while it suffers from rapid increasing after 40 users for PF algorithm. If two Pico cells are added to the aforementioned system, a similar performance is shown, but the delay value is decreased. In addition, the threshold of PF is shifted at 60 users instead of 40 users in the single Macro case. To compare MLWDF and EXP/PF in both scenarios, a certain point in Figures.4.8 and 4.9 could be explained. For example at 60 users, in a single Macro cell the delay value is 50 ms while the value is 20 ms in the Macro with 2 Picos. As a consequence, for MLWDF and EXP/PF, the delay value with two Picos is approximately half the delay value without Pico cells. One of the purposes of HetNets is to enhance the latency, and this is viewed in a practical simulation illustrated in Figure.4.9. However, the delay manifests lower values (nearly 10 times lower) in the scenario of the single cell with 2 Picos using PF scheme. 76

88 Figure 4.8 Packet Delay of Video Flows (single Macro cell) Figure 4.9 Packet Delay of Video Flows (Macro with 2 Picos) Fairness Index When the number of users increases in single Macro cell more than 30, the fairness index of all simulated algorithms is deviated down of the value 1. At 40 users, PF experiments further deviation close to value 0.8 compare with other algorithms that they are around 0.9 as seen in Figure The fairness index behaves similarly in the scenario of Macro with 2 Picos as shown in Figure However, the PF shows a minor different in which at 50 users it starts to decline to get the value

89 Figure 4.10 Fairness Index of Video Flows [15] Figure 4.11 Fairness Index of Video Flows Macro with 2 Picos 4.5. Simulation.2-Single Macro Cell with two Pico Cells (Different Speed Comparison) This simulation has been done to compare between telecommunication systems that compromise of a Macro cell with two small low power Pico cells where the users are moving in 3 Km/hand in one scenario and 120 Km/h in the other scenario. 78

90 Simulation.2 Environment Similar simulation environment that have been mentioned in Simulation1 is applied in this simulation. LTE-Sim platform is used with the added code that demonstrates the new scenario of Macro with two Picos. The parameters in the Table 4.2 are applied, but the speed is changed to be 120 Km/h instead of 3 Km/h. Similar Pico cells positions in the Macro cell as seen in Figure.4.3 is considered. This distribution is more likely close to the normal Pico positions in the practical networks, in which operators locate Pico cells in Macro boundary to extend the coverage, increase capacity and boost the throughput. Useless Pico cells will be if their base stations near the Macro enb position, that is, the Macro enb already serves the UE. Handover is activated, and 3GPP urban Macro cell propagation loss model has been performed as aforementioned in simulation1. Initially, each HetNets cell has a certain number of users who can be handed over during the simulation between Macro and Picos and vice versa. In this simulation, number of users increase equally in all cells. However, theoretically Pico cell could serve UEs simultaneously [17]. Each user has 50% video flows and 50% VoIP. The outcomes are based on the video traffic Simulation.2 Results It could predict the simulation results since high speed degrades the system behaviour the users connectivity with system s cells will be worse. The degradation is shown as reducing overall system throughput while increasing PLR, latency and fairness Throughput As seen in Figure.4.12, the dashed lines denote PF, MLWDF and EXP/PF in the case of 120 Km/h speed. It is obvious that the system is degraded due to the speed increment for each user. For example, the point 60 users has 16 Mbps throughput value if the mobility speed is 120 km/h while in the case of 3 Km/h the same point has more than 25 Mbps for all scheduling schemes. However, if it is compared with Figure.4.5 (throughput in Macro cell only at speed of 3 Km/h), the system manifests better performance although the speed is 120 Km/h. This is due to the positive impact of adding two Pico cells. As an example, the maximum throughput value viewed in Figure.4.5 is almost 12 Mbps due to the effect that there are no Pico cells. 79

91 Figure 4.12 Throughput of Video in Macro with 2 Picos (3 Km/h and 120 Km/h speed) Packet Loss Ratio (PLR) The PLR value is highly affected by the speed of mobility. Packets are likely to suffer of errors and could be dropped while the speed is increased due to the fact that the connectivity with the base stations gets worse. As seen in Figure.4.13, the PLR value is higher in the scenario of 120 Km/h speed, in which the higher PLR, the worse system performance. Considering point 60 of users, the average PLR values for all scheduling algorithms is 0.33 while the speed is 120 Km/h. Same point at speed 3 Km/h illustrates lower PLR values for all schedulers, for example, PLR of PF is 0.14 which is almost half the value of 0.33 in the 120 Km/h scenario. The scheduling schemes are performing similarly. However, MLWDF and EXP/PF outperform PF in both scenarios. Figure 4.13 PLR of Video in Macro with 2 Picos (3 Km/h and 120 Km/h speed) 80

92 Delay Figure.4.14 shows the delay of the system in both when the speed is 120 Km/h (the dashed line) and the speed 3 Km/h (the straight line). The delay is higher in the scenario of 120 Km/h especially with the PF scheme. Maximum delay in the case of 3 Km/h is 20 ms as experimented in this simulation.exp/pf and MLWDF have similar behaviours in the both scenarios although the delay in the case of 120 Km/h is almost double the value of that in the case of speed of walking. For instance, at 60 users in the 120 Km/h speed, the EXP/PF and MLWDF has a delay of 44 ms while in the other case the delay is 20 ms which is nearly half the value of 44 ms. Figure 4.14 Delay of Video in Macro with 2 Picos (3 Km/h and 120 Km/h speed) Fairness Index The system provides fairness values similar to those in the simulation1.pf is outperformed by MLWDF and EXP/PF where it shows more decline down the value of one as the number of users increases. The speed has an impact on the fairness values. As it is seen in Figure.4.15, the 120 Km/h scenario enhances the fairness slightly than the 3 Km/h scenario for all scheduling schemes. This gives a good indication that using HetNets (Macro with Pico cells) with high speed mobility not only keeps the system performing similarly, but also could enhance the fairness index. However, PLR, delay and throughput have lower values using high speed mobility, thereby; the overall system is degraded when the users are moving fast. 81

93 Figure 4.15 Fairness Index in Macro with 2 Picos (3 Km/h and 120 Km/h speed) 4.6. Simulation.3- Single Macro Cell with Increasing Pico Cells Third Simulation has been conducted to compare between different scenarios of a single Macro cell telecommunication system by adding more Pico cells. The increment is constant; that is, in each case two more Pico cells are added in new positions at the Macro cell edge Simulation.3 Environment Same simulation environment has been used for the rest of the study. However, further updated code has been used in LTE-Sim simulation platform to perform this analysis. The simulation is based on a scenario of a single Macro cell with 2, 4, 6, 8 and 10 small Pico cells. Each Pico transmits 30 dbm of power while the Macro cell transmits 49 dbm. Table 4.2 parameters are set up in all scenarios to maintain same values for the system while increasing the Pico cells by factor of 2. This ensures that all new results of the system performance come through the factor of adding Pico cells only. However, flow duration and simulation time have been modified to 30 and 40 respectively. All other aforementioned simulation environment elements are similarly utilised such as 3GPP urban Macro cell path loss and handover activation. Users at Pico cells are not equal to the Macro cell users. Pico cannot serve more than 30 users [17] users, thereby, after 30 users there are no more users could be added. Nevertheless, all cells start with 10 users and increase by factor of 10. Flows are equally divided into Video and VoIP flows, in which each of them is 50% of the total system traffics. In this simulation, there are two sides have to be considered. One of them is increasing the number of Pico cell gradually by factor of 2, and the other one is increasing the number of users by the factor of 10. Because of this, 3D graphs are 82

94 used to represent the system performance besides using 2D graphs to study the system behaviour in different scheduling schemes. The distribution of the Pico cells within Macro boundary follows Figure Table 4.3 summarises the (x,y) values for each position from the simulation outcomes. Figure 4.16 Applied HetNets (Macro with Multiple Picos Scenarios) Macro (x,y) 2 Pico cells 4 Pico cells(x,y) 6 Pico cells (x,y) (x,y) id 0, position: 0, 0 id 1, position: 1000, 0 id 2, position: , 1.5 id 1, position: 500.4, id 2, position: -499, id 3, position: , id 4, position: 497.6, id 1, position: 1000, 0 id 2, position: 500.4, id 3, position: -499, id 4, position: , 1.5 id 5, position: , id 6, position: 497.6, Pico cells(x,y) 10 Pico cells(x,y) id 1, position: 1000, 0 id 2, position: 707.3, id 3, position: 0.79, 1000 id 4, position: , id 5, position: , 1.5 id 6, position: , id 7, position: -2.3, id 8, position: 705.1,-709 id 1, position: 1000, 0 id 2, position: 809.2, id 3, position: 309.6, id 4, position: , id 5, position: , id 6, position: , 1.5 id 7, position: , id 8, position: , id 9, position: 306.5, id 10, position: 807.3, Table 4.3 Pico Cells Positions in meters into the Macro Cell (Radius 1 Km) 83

95 Simulation.3 Results As mentioned before, adding more Picos more likely enhances the system performance. This is proven in this simulation demonstrated through the throughput, PLR, delay and fairness. Adding 2 extra Pico cell improves the overall system throughput with a certain value which cannot be normalized.this is because number of reasons such as increasing the number of users, increasing the effect of inter-cell interference while the number of users increases and simulation parameters including power value, simulation time, flow duration affect that while the number of users increases Throughput The average overall system throughput for all scenarios is seen in Figure Adding 2 Pico cells provides almost constant gain in all applied scheduling schemes. To explain that, PF algorithm (Table.4.4 in 5 cases) is taken as an example. Vertically, the number of users increases by factor of 10, and by taking the cases (10 users) to (30 users) the gain value is 3.39 Mbps due to UEs increments. For example, at 20 users the average value of for all scenarios is 6.78 Mbps that equals the value of 3.39 at 10 users plus the 3.39 Mbps gain. The difference of average gain between 30 users point and 20 users point is also 3.39 Mbps that proves adding 10 users increases the throughput in constant value of the gain while adding 2 Pico cells to the system. After 30 users, the Picos cannot serve more UEs, and the gain will continue at almost the same value of the gain at 30 users, which is 10 Mbps as average. However, adding more users to the system (only Macro users) boosts the throughput slightly with nearly 1 Mbps due to the effect of scheduling algorithms only. Horizontally, moving up from 2 to 10 Pico scenarios the throughput value increases by average gain of 10 Mbps. Figure.4.18 gives another view of the gain and shows clearly the system performance in each case. All aforementioned values of the gain are not constant, and they are based on the simulation parameters and system environment. 84

96 PF Picos (Yaxes) Throughput( Z-axes) Average 2 Picos gain [Mbps] Mbps increment Users (Xaxes) Mbps increment Mbps increment Mbps increment Mbps increment Mbps increment Mbps increment Mbps increment Table 4.4 Throughput Gain Values and An Average of The Values Figure 4.17 Throughput Gain of Video traffic in Macro with 2-10 Picos Scenarios 85

97 Figure.4.18 Throughput Gain of Video traffic in Macro with 2-10 Picos Scenarios Packet Loss Ratio (PLR) Although the number of users increases that means the PLR value increases accordingly, the PLR values start getting down as the number of Pico cells raises. Adding more Picos can be equivalent to the PLR increment due to more users is added to the system. For instance, PF algorithm is the lower performance than other algorithms in most of the cases, in which the PLR starts going up while the system is charged with more users. However, as seen in Figure.4.19 at the 50 users, PF with 8 Pico cells case has the same PLR value of other schemes with 2 Pico cells scenarios. This enforces the idea of equivalent; that is, adding more Picos enhances system PLR of PF bringing it back to the value where other algorithms are in. Figure.4.20 gives 3D view of the PLR behaviour in the dimension of adding more users and the dimension of adding more Pico cells. 86

98 Figure 4.19 PLR Video traffic Comparison in Macro with 2-10 Picos Scenarios Figure 4.20 PLR of Video traffic in Macro with 2-10 Picos Scenarios 87

99 Delay Delay follows similar behaviour to the PLR. While the number of users increases, the delay gets higher. Reverse of that, while Pico cells are added to the system, the delay becomes lower. Table 4.5 is an example of PF delay values that are obtained from the simulation to draw Figure.4.21 and Similarly, MLWDF and EXP/PF have been drawn. PF has the higher delay values in all scenarios and the highest value in the 2 Pico cell case while the number of users at the maximum in this simulation. It is easily to notice that PF starts decreasing while the number of Pico cells increases. Similar performance for all algorithms in all scenarios is viewed in Figure MLWDF and EXP/PF analogy the stairs, in which lower delay values are at 10 Pico cells scenario climbing up to the higher at 2 Pico cells scenario. Picos (Y-axes) Delay (Zaxes) Users (Xaxes) Table 4.5 PF Throughput Gain Values and An Average of The Values 88

100 Figure 4.21 Delay of Video traffic Comparison in Macro with 2-10 Picos Scenarios Figure 4.22 Comparison Delay of Video traffic in Macro with 2-10 Picos Scenarios 89

101 Fairness Index The fairness index has to be closer to the value of one. Adding more users affects this value that starts slope down. Figure.4.23 shows the fairness index of the system that has 2 to 10 Pico cells. The value slightly declines from one for the MLWDF and EXP/PF while the PF suffers further drop from the value of one. Adding more Pico cells has no effect on the fairness as shown from the Figure.4.23 and Figure.4.24.Compare with the same system without Pico cells (as seen in Figure.4.10), the system has similar behaviour although adding more cell slightly enhances the overall system fairness value. From the values of all algorithms, the worst scenario of the fairness index is at 8 Pico cells in which PF shows the lowest fairness value when the number of users is 80 and the scenario is 8 Pico cells. MLWDF and EXP/PF also suffers further drop in the same point as aforementioned in PF. Figure.4.24 illustrates that. Figure 4.23 Fairness Index in Macro with 2-10 Picos Scenarios 90

102 Figure 4.24 Fairness Index in Macro with 2-10 Picos Scenarios 4.7. Conclusion This chapter investigates scheduling algorithms that are developed to enhance the LTE network performance by sharing radio resources fairly among users utilizing all available resources. These algorithms depend on traffic class and number of users, hence; different outcomes are presented for each algorithm. To further boost the overall system performance, this study uses heterogeneous networks concept by adding small cells (initially 2 Pico cells). This enhancement is experienced through a throughput, PLR, delay and fairness. In the throughput the system gains more data rate while in PLR the system suffers less packet loss values. Moreover, delay is decreased and fairness stays similar. Approximately from the simulation1 outcomes, the overall system performance is as follows: throughput is duplicated or nearly tripled relaying on the number of users, the PLR is almost quartered, the delay is reduced 10 times (PF case) and changed to be a half value (MLWDF/EXP cases), and the fairness stays closer to value of 1. On the other hand, high speed mobility in simulation2 degrades the overall system performance although the system appears better fairness index. 91

103 Lastly, as the number of small cell increases as determined in simulation3, the system manifests more enhancements as seen in 2D and 3D graphs for throughput, PLR, delay and fairness. However, it is expected that a saturation state will be reached after a certain point of the number of Pico cells and the number of users. The reason behind that is the inter-cell interference will limit the performance since the same carrier frequency is used in all system s cells. Considering all scenarios, MLWDF manifests the best performance for video flows followed by EXP/PF. Further enhancement could be applied in future papers such as almost blank subframes (ABS), enhanced inter-cell interference cancelation (eicic), cell range extension CRE concepts. and using Carrier Aggregation (CA) and CoMP within HetNets. References [1] H. A. M. Ramli, R. Basukala, K. Sandrasegaran, and R. Patachaianand, "Performance of well known packet scheduling algorithms in the downlink 3GPP LTE system," in Communications (MICC), 2009 IEEE 9th Malaysia International Conference on, 2009, pp [2] Seung June Yi, S.C., Young Dae Lee, Sung Jun Park, Sung Hoon Jung 2012, Radio Protocols for LTE and LTE-Advanced. [3] [1] B. Liu, H. Tian, and L. Xu, "An efficient downlink packet scheduling algorithm for real time traffics in LTE systems," in Consumer Communications and Networking Conference (CCNC), 2013 IEEE, 2013, pp [4] A. Jalali, R. Padovani, and R. Pankaj, "Data throughput of CDMA-HDR a high efficiency-high data rate personal communication wireless system," in Vehicular Technology Conference Proceedings, VTC 2000-Spring Tokyo IEEE 51st, 2000, pp [5] M. Andrews, K. Kumaran, K. Ramanan, A. Stolyar, P. Whiting, and R. Vijayakumar, "Providing quality of service over a shared wireless link," Communications Magazine, IEEE, vol. 39, pp , [6] J.-H. Rhee, J. M. Holtzman, and D. K. Kim, "Performance analysis of the adaptive EXP/PF channel scheduler in an AMC/TDM system," Communications Letters, IEEE, vol. 8, pp , [7] J. Zyren and W. McCoy, "Overview of the 3GPP long term evolution physical layer," Freescale Semiconductor, Inc., white paper, [8] B. Riyaj, M. R. H. Adibah, and S. Kumbesan, "Performance analysis of EXP/PF and M-LWDF in downlink 3GPP LTE system," [9] X. Qiu and K. Chawla, "On the performance of adaptive modulation in cellular systems," Communications, IEEE Transactions on, vol. 47, pp , [10] S. C. Nguyen, K. Sandrasegaran, and F. M. J. Madani, "Modeling and simulation of packet scheduling in the downlink LTE-advanced," in Communications (APCC), th Asia-Pacific Conference on, 2011, pp [11] A. Alfayly, I.-H. Mkwawa, L. Sun, and E. Ifeachor, "QoE-based performance evaluation of scheduling algorithms over LTE," in Globecom Workshops (GC Wkshps), 2012 IEEE, 2012, pp

104 [12] G. Piro, L. A. Grieco, G. Boggia, F. Capozzi, and P. Camarda, "Simulating LTE cellular systems: an opensource framework," Vehicular Technology, IEEE Transactions on, vol. 60, pp , [13] M. Iturralde, T. Ali Yahiya, A. Wei, and A. Beylot, "Resource allocation using shapley value in LTE networks," in Personal Indoor and Mobile Radio Communications (PIMRC), 2011 IEEE 22nd International Symposium on, 2011, pp [14] R. Jain, D.-M. Chiu, and W. R. Hawe, A quantitative measure of fairness and discrimination for resource allocation in shared computer system: Eastern Research Laboratory, Digital Equipment Corporation, [15] AL-Jaradat, Huthaifa 2013, On the Performance of PF, MLWDF and EXP/PF algorithms in LTE. [16] Holma H, Toskala A 2012, LTE-Advanced 3GPP Solution for IMT-Advanced. [17] Hu, Rose Qingyang Qian, Yi 2013, Comparison Femto cell and Pico cell key features, Heterogeneous Cellular Networks (2nd Edition). 93

105 Research Proposal LTE-Advance Enhancement Using CA and CoMP within HetNets By Haider Al Kim November 2014 Supervisor: Dr. Kumbesan Sandrasegaran 1

106 Introduction The telecommunication networks have rapidly been updated since 1980 s where 1G of mobile telecommunication was proposed. Low traffic capacity, poor call quality and higher power usage are characteristics of 1G network. Available network resources have to be utilized efficiently to increase cell capacity, coverage and satisfy Quality of service QoS requirements. Most recent trends of researchers are designing models that can meet users expectations. Although modern networks have been designed carefully to fulfill the requirements needed by the end users; significant challenges have emerged. There is a limited bandwidth that has to be used sufficiently. RRM and air interface techniques have become the research interest fields for the researchers. 2

107 Contents Mobile Telecommunications Trends Technology Review Focus Area Statement of the problem Methodology Research Timeline Bibliography 3

108 Mobile Telecommunications Trends Mobile networks have already grown rapidly passing some remarkable signs. Rapid computer technologies have led to short-period mobile evolution which is aimed to meet the increment of higher data rate requirements and satisfy a certain agreed QoS. This emulation is shown in Figure.1.3GPP standardizes the most important demands that have to be met in order to cross the mobile system to the new generation that is 4G. 4G fulfills the IMT-Advance constraints which are agreed to increase the user expectation. At the top of the hierarchy, LTE-A, a subset of Rel- 10, has been proposed with significant challenges. The term LTE-A also refers to 4G although informally the term 4G is used for WiMAXTM.4G is commercially used as a term by some operators to describe HSPA evolution. International Telecommunications Union Radio communication Sector (ITU-R) has involved in the development of the proposed system by 3GPP introducing the Release 10 or what so-called recently LTE-A. The ITU-R involvement in specifying the LTE-A requirements has complicated the process of setting up Rel-10. Although Rel-8 could meet most of the 4G requirements, LTE-A has other features that could not be satisfied by LTE. These LTE-A-based requirements are higher bandwidth coming from carriers aggregating (CA), and higher efficiency could be conducted using higher uplink multiple access technologies and enhanced multi-in-multi-out MIMO antennas. Further enhancements could be as essential parts of LTE-A network, but they do not have to be LTE-A requirements. These features are: - Support for heterogeneous networks (HetNets) and Relaying - Coordinated multipoint transmission/reception (CoMP) - LTE self-optimizing network (SON) enhancements - Mobility enhancements for Home enhanced-node-b (HeNB) - RF requirements for fixed wireless customer premises equipment (CPE) 4

109 Figure.1 Wireless evolution and beyond: Technology Review Different access technologies have been introduced with each mobile generation network trying to address the problems in the previous versions. FDMA network that is an access method used in the first generation of the mobile telecom-systems has been modified presenting what socalled orthogonal FDMA that is adopted to be applied in the new mobile generations. Technically, FDMA is a divided available system bandwidth into non-overlapping frequencies. The main usage of FDMA was for analogue systems that have been changed later to modern digitized systems. For the second generation 2G, TDMA is applied as an access technology. That is; the division is based on the time intervals for each call. Analogue-to-digital converters are used in TDMA to produce digital signals constructing consolidated digital stream that can be carried on a single radio channel. While the telecommunication systems continued development, modern access methods have been adopted. CDMA is the new access technique that is used with 3 rd generation (3G). It is considered a very efficient method to avoid overlapping of FDMA and the limitation of the time interval of TDMA. Using the code to separate between conversations is what CDMA has brought. More data rate is required in the recent days due to the demand of high-resolution video streams and high-quality voice conversations.3gpp provided the standards of long-term evaluation system LTE. LTE uses most recent technologies in mobile networks. 5

110 One of them is Orthogonal Frequency Division Multiple Access technology OFDMA that is utilized mainly to minimize interference effect for overlapped frequencies. Another technology is MIMO that plays a major role in the LTE system performance that further enhanced to propose advanced MIMO. LTE with these technologies provides higher efficiency of the spectrum, lower delay and seamless handover, thereby, performs better than the previous systems. OFDMA: it is one of a key element in LTE network which is basically used to robust the resistance to multipath fading and interference as well as it is considered as a digital signal processing techniques. It guarantees little updating to the existed air interface while flexible deploying over available frequencies. It also provides an average value of the inter-cell interference caused by neighboring cells and an average value for intra-cell interference caused by overlapped frequencies. By spreading the carriers over the available spectrum, OFDMA provides frequency diversity and excellent coverage. It uses large, narrow band (180 khz) subcarriers for multi-carrier transmission to carry data. Figure.2 shows the basic LTE downlink physical resource where OFDM symbols are grouped into resource blocks. Figure.2 Basic LTE downlink physical resource using OFDMA: - MIMO/Advanced MIMO: it is antenna structure that is adopted to robust the data rate and maximize the performance in LTE-A. The expected LTE-A MIMO is 8x8 downlink antenna configuration while (4x4) antenna configuration is proposed to be utilized in the uplink direction. It is one of the suggested smart antenna technologies. The significant benefit of MIMO is that it provides higher data rate without the need to increase the 6

111 bandwidth or the transmission power. This can be conducted by spreading the used power of the transmission among the antennas to enhance the spectral efficiency by obtaining array gain. - Relaying and Heterogeneous networks (Macro with Pico or Femto cells): relay, Pico and Femto are small low power nodes that inserted within or on the edge of the large mobile cells to enhance the throughput and increase the capacity and coverage. A relay is slightly different since it protocol structure recently reaches layer 3 (router more than to be a repeater only). However, the main purpose of it is that retransmitting the received signal without modifying to the far end that it is out of the main coverage area of large cell. Small cells and relay are considered an effective air interface enhancement methods which require lower cost and little modifying to the existing mobile networks. - Carrier Aggregation CA: in LTE, single carrier is allocated to the LTE user. When LTE- A is proposed, the demand for more data rate that has to be provided to the LTE-A user is studied. Hence, CA is applied to increase the bandwidth, thereby, it increases the bit rate. On the other hand, backward compatibility exists with legacy schemes Rel8/9 that means they can co-exist with LTE-A where the CA is based on Rel8/9 carrier component. The maximum LTE available bandwidth is 20 MHz. By aggregating 5 of 20 MHz, the new LTE-A bandwidth is 100 MHz that provides a higher rate of throughput. Recent studies are proposed different algorithms to aggregate carrier in LTE-A based on that if the carriers are on the same frequency band (contiguous or non- contiguous) or cross-carrier frequency (non- contiguous only). - Coordinated Multi-Point (CoMP) Transmission/Reception is a mechanism, in which a number of geographically separated enbs are cooperated to serve one user in the network in order to improve the performance the users in the covered areas. The suggested method of connecting these enbs is using high speed dedicated connections, for example, microwave links or optical fiber. The inter-cell interference impact is affirmatively minimized using CoMP in both the downlink and uplink directions [4]. 7

112 Focus Area Deploying Heterogeneous networks (HetNets) approach is one of the probable key features of future LTE-A networks. There are different methods to apply HetNets in wireless systems. Using separate frequency band in a small cell from the frequency in a large cell is to avoid the interference. This method is call dedicated carrier HetNets. There is a drawback of using dedicated carrier, in which the probability of inefficient usage of the frequency band exists. Moreover, intra-frequency handover is required while the user is moving between HetNets cells. The most common approach of HetNets (Macro with small cells) is applying the same frequency band in all cells with HetNets. This increases the inter-cell interference, but it could introduce higher spectral efficiency. However, careful interworking between HetNets cells is vital in the scenario of using the same frequency in all HetNets cells. That is; it requires a centralized node to control the HetNets cells, which is called Remote Radio Heads (RRH). According to NTT DoCoMo [5], new base station is proposed that using advanced centralized Radio Access Network structure. Once this centralized architecture is deployed using multiple bands, it is possible to use carrier aggregation in Rel-10 as an extension of a base station. CA could combine contagious or noncontiguous frequency bands to create LTE required bandwidth. Normally, CA is conducted on the same cell using the available carriers. However, CA is enhanced to be involved within multi-cells using the centralized node. With HetNets, a small enb has been introduced as a low-cost base station with a reduced transmission power. Rel-11 provided a multi-carrier aggregation using the timing advance enhanced uplink power control [6]. Figure.3 illustrates HetNets with primary cell (large-macro cell) and secondary cell (small Pico or Femto). In this scenario, the Large cell is responsible for providing control signaling, system information and limited data transmission while the small cell is responsible for providing the required high data rate. It is beneficial in both cases, CA with dedicated frequency or co-channel. Due to some drawbacks of this CA in HetNets such as users terminal compatibility a with multi-carrier aggregation of Multiple 8

113 Timing Advance, another method that does not depend on the centralized node can be used. That is, large and small cells can operate with own control signaling for both layered frequencies. This approach requires enhanced interference management such as ICIC. However, ICIC is limited to the PDSCH data; therefore, it requires new solutions to separate the control channels. PDCCH is used to provide full control channel protection especially if it is used as crosscarrier scheduling. In this case, the interference will be at its minimum value if the small cell does not use the PDCCH. Figure.4 shows the concept of using PDCCH in co-channel HetNets. Figure.3 Multi-Carrier Aggregation in LTE-A HetNets [6] Figure.4 CA in co-channel scenario in LTE-A HetNets [6] 9

114 Statement of the problem Compare with Time Domain Interference Coordination using ABS (Almost Blank subframe) in eicic, the aforementioned Frequency Domain Interference Coordination using PDCCH cross-carrier scheduling has some benefits. Using eicic introduces more complexity in the network, for example, signaling and measurements are more likely to be higher in a co-channel deployment. However, there is a probability that Macro cell passes out the PDCCH since it uses cross-carrier scheduling. Moreover, to support MIMO, enhanced PDCCH is required (epdcch) that is already addressed in 3GPP Rel-11 [6]. The current deployment of small cell and relays is using the ideal backhaul which depends on a centralized architecture and supports easily CA and CoMP operations [4]. Studies are proposed new methods of deploying small cell with non-ideal backhaul. Currently in Rel-11, it could aggregate two TDD carriers with different configuration of uplink and downlink. Such configuration requires UE has an ability to transmit and receive in parallel. This could lead to that such system is similar to FDD. Hence, operators could combine FDD and TDD spectrums in one solution. For HetNets, a clustered TDD small cell deployment could be possible as shown in Figure.5. By separating these clusters, a dynamic adjacent of uplink / downlink frame structure could be possible relying on the need of local traffic in the small cells. 10

115 Figure.5 Dynamic TDD in LTE-A HetNets [6] Using HetNet with CA and CoMP is the key evolution of future networks. Investigating using HetNets small cell as corporative cells to apply CA and CoMP techniques is a major interest in this proposal. Intensive most recent papers will be reviewed and studied that address and discuss the current challenges as motioned before ideal backhaul as a centralized architecture where CA and CoMP rely on. Figure.6 shows the suggested scenario for future HetNets. Hence, HetNets deployment not only supports large cell to serve some users who suffer from bad connectivity with Macro cell, but also could be utilized simultaneously as corporative cells to apply CA and CoMP concept. It could further increase throughput, coverage and capacity besides reducing the latency. More enhancements for small cell will be considered, such as using 256 QAM to enhance spectral efficiency, enhanced inter-frequency measurement and enhance interference coordination to improve small cell operation [6]. 11

116 Figure.6 CA with ideal and non-ideal backhaul in the suggested HetNets [6] 12

117 Methodology Technology Review Searching State of the art scientific papers Identifying existed problem/technology limitation Preparing developing model/writing paper Code writing and Simulation implementation Results/ Conclusions Writing Thesis 13

118 Research Timeline Task Technology Review Searching State of the art scientific papers Identifying existed problem/technology limitation Preparing developing model/writing paper 1st Semester First Year Second Year Third Year Forth Year 2nd Semester 3rd Semester 4th Semester 5th Semester 6th Semester 7th Semester 8th Semester Code writing and Simulation implementation Results/ Conclusions Writing Thesis /Extension if required 14

119 Bibliography [1] A. T. Moray Rumney, LTE and the Evolution to 4G Wireless: Design and Measurement Challenges, Second Edition. John Wiley and Sons, Ltd, [2] A. Toskala and H. Holma, WCDMA for UMTS HSPA Evolution and LTE, Fourth Edition. John Wiley and Sons, Ltd, [3] D. K. Sandrasegaran, \Lecture note, lte radio resource management," tech. rep., University of Technology, Sydney. [4] I. F. Akyildiz, D. M. Gutierrez-Estevez, and E. C. Reyes, "The evolution to 4G cellular systems: LTE-Advanced," Physical Communication, vol. 3, pp , [5] Press Release NTT DoCoMo TOKYO, JAPAN, February 21, 2013: DOCOMO to Develop Next generation Base Stations Utilizing Advanced C- RAN Architecture for LTE-Advanced [6] Eiko Seidel, LTE-A HetNets using Carrier Aggregation, NoMoR Research GmbH, Munich, Germany, June

120 Appendix - LTE-Sim Macro with Pico Applied Code - Published Paper (IJWMN)

121 APPLIED CODE 1- Simulations Parameters Type (Shell Code,.sh) 2- Simulations Parameters Type (C++,.h) 1

122 3- Single Macro with Multiple Pico Cells - Cells Positions Part of the Code Type (C++,.h) 4- Single Macro with Multiple Pico Cells Create Pico Cells Part of the Code Type (C++,.h) 2

123 5- Main LTE-Sim Execution File includes Single Macro with Multi Pico Passing Parameters from the (.sh) file (C++,.cpp) 6- An Example of The First Part of Simulation File Outcomes 3

124 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October 2014 MACRO WITH PICO CELLS (HETNETS) SYSTEM BEHAVIOUR USING WELL-KNOWN SCHEDULING ALGORITHMS Haider Al Kim 1, Shouman Barua 2, Pantha Ghosal 2 and Kumbesan Sandrasegaran 2 1 Faculty of Engineering and Information Technology, University of Technology Sydney, Australia {Haider.A.AlKim} {shouman.barua, pantha.ghosal, kumbesan.sandrasegaran} Abstract This paper demonstrates the concept of using Heterogeneous networks (HetNets) to improve Long Term Evolution (LTE) system by introducing the LTE Advance (LTE-A). The type of HetNets that has been chosen for this study is Macro with Pico cells. Comparing the system performance with and without Pico cells has clearly illustrated using three well-known scheduling algorithms (Proportional Fair PF, Maximum Largest Weighted Delay First MLWDF and Exponential/Proportional Fair EXP/PF). The system is judged based on throughput, Packet Loss Ratio PLR, delay and fairness.. A simulation platform called LTE-Sim has been used to collect the data and produce the paper s outcomes and graphs. The results prove that adding Pico cells enhances the overall system performance. From the simulation outcomes, the overall system performance is as follows: throughput is duplicated or tripled based on the number of users, the PLR is almost quartered, the delay is nearly reduced ten times (PF case) and changed to be a half (MLWDF/EXP cases), and the fairness stays closer to value of 1. It is considered an efficient and cost effective way to increase the throughput, coverage and reduce the latency. Keywords HetNets, LTE &LTE-A, Macro, Pico, Scheduling algorithms & LTE-Sim 1. INTRODUCTION In the Long Term Evolution so-called LTE, the requirements for larger coverage area, more capacity, and high data rate and low latency have led to search for cost-effective solutions to meet these demands. Hence, the development in the telecommunication networks has adopted different directions to enhance the LTE system taking into account the International Mobile Telecommunications (IMT-2000) standards that have to be satisfied [1]. Network-based technologies such as Multiple Input and Multiple Output MIMO/ advanced MIMO and Transmission/Reception Coordinated Multi-Point CoMP are LTE enhancements that introduce LTE Advance (LTE-A). Other less cost enhancements based on air interfaces are proposed, such as improving spectral efficiency involving using Heterogeneous networks (HetNets). HetNets are small and less power cells within the main macro cells with different access technologies to close up the network to the end users and increase their expectation [16].According to [2], there are two main practical HetNets classes: Macro with Femto and Macro with Pico. Femto and Pico are the small and less power cells. To save the cost, operators use the same carrier frequency in the large and small cells which, on the other hand, proposes interference challenges. Figure 1 gives the main concept of HetNets. To clarify, user in LTE is well-known as a UE. DOI : /ijwmn

125 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October 2014 Figure.1 an example of HetNets In LTE and LTE-A, the element that is responsible for Radio Resources Management (RRM) is enhanced Node Base station (so-called enb). The enb does all required management including Packet Scheduling (PS) which is the focus in the paper. PS can guarantee the agreed quality of service demands (QoS) because it is responsible for the best and effective utilizing of the affordable radio resources and in charge of data packets transmission of the users[3]. 3rd Generation Partnership Project (3GPP) has left the scheduling algorithms to be vendor specific according to user s requirements and network capability. Therefore, various PS algorithms have been proposed depending on the traffic sorts and provided services. PF, MLWDF and EXP/PF algorithms [4][5][6] are used in this paper to study and compare between the system behaviours in HetNets (single Macro with 2 Pico cells) using these three types of algorithms. Scheduling algorithms ensure that QoS requirements have been met. This can be conducted by prioritizing each link between the enb and the users, the higher priority connection the first handled in the enb. This paper is organized as follows. Section II discusses the downlink system model of LTE. The followed section (III) describes in more details packet scheduling algorithms, while Section IV present simulation environment. Section V shows the outcomes of the simulation. Finally, conclusion is given in Section VI. 2. DOWNLINK SYSTEM MODEL OF LTE The basic element in the downlink direction of the LTE networks is called Resource Block (RB).Each UE is allocated certain number of resource blocks according to its status, the traffic type and QoS requirements. It could define the RB in both frequency domain and time domain. In the time domain, it comprises single (0.5 ms) time slot involving 7 symbols of OFDMA (orthogonal frequency division multiple access). In the frequency domain, on the other hand, it consists of twelve 15 khz contiguous subcarriers resulting in 180 khz as a total RB bandwidth [7]. As aforementioned before, the enb is responsible for PS and other RRM mechanisms. The bandwidth that is used in this study is 10 MHz considering the inter-cell interference is existed. The period that enb performs new packet scheduling operation is the Transmission Time Interval (TTI). 110

126 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October 2014 TTI is 1 ms that mean the users are allocated 2 contiguous radio resource blocks (2RBs). The scheduling decision in the serving enb is made based on the uplink direction reports come from the UEs at each transmission time interval. The reports comprise the channel conditions on each RB, such as signal to noise ratio (SNR). The serving enb uses the SNR value involved in the reports to specify the DL data rate for each served UE in each TTI. For example, how many bits per 2 contiguous RBs [8]. The data rate for user i at j sub-carrier on RB and at t time can be determined by using equation (1) as proposed in [9]. (1) A = B = C = D = rgg The number of bits per symbol is A. The number of symbols per slot is B. While C represents how many slots per TTI, D clarifies how many sub-carriers per RB. Table 1 summarizes the mapping between SNR values and their associated data rates. Table 1. Mapping between instantaneous downlink SNR and data rate Minimum SNR Modulation and Data Rate Level (db) coding (Kbps) 1.7 QPSK (1/2) QPSK (2/3) QPSK (3/4) QAM (1/2) QAM (2/3) QAM (3/4) QAM (2/3) QAM (3/4) 756 Upon the packets reach the enb, they are buffered in enb in a specific container allocated for each active UE. Moreover, the buffered packets are assigned a time stamp to ensure that they will be scheduled or dropped before the scheduling time interval is expired, and then using First-In-First-Out (FIFO) method they are transmitted to the users in the downlink direction. To explain the scheduling operation, PS manager (is a part of enb functionalities) at each TTI priorities and classifies the arriving users packets according to preconfigured scheduling algorithm. Scheduling decision is made based on different scheduling criteria that have been used in various algorithms. For example channel condition, service type, Head-of-Line (HOL) packet delay, buffer status, and so on so forth. One or more RBs could be allocated to the selected user for transmission with the highest priority. Figure 2 shows the packet scheduler in the downlink direction at enb. 111

127 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October 2014 Figure.2 Downlink Packet Scheduler of the 3GPP LTE System [10] 3. PACKET SCHEDULLING ALGORITHMS The efficient radio resource utilization and ensuring fairness among connected users, as well as satisfying QoS requirements, are the main purposes of using PS algorithms [11].The PS algorithms that have been used in this study are : Proportional Fair (PF) algorithm, Maximum-Largest Weighted Delay First (MLWDF or ML) and the Exponential/Proportional Fair (EXP/PF or EXP) algorithm. It should be noted that these algorithms are used Proportional Fair (PF) Algorithm For non-real time traffic, the PF was proposed which is used in a Code Division Multiple Access- High Data Rate (CDMA-HDR) system in order to support Non-Real Time (NRT) traffic. In this algorithm, the trade-off between fairness among users and the total system throughput is presented. This is, before allocating RBs, it considers the conditions of the channel and the past data rate. Any scheduled user in PF algorithm is assigned radio resources if it maximizes the metric k that calculated as the ratio of reachable data rate of user i at time t and average data rate of the same user at the same time interval t: where; (2) (3) is the window size used to update the past data rates values in which the PF algorithm maximizes the fairness and throughput for any scheduled user. Unless user i is selected for transmission at, =

128 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October Maximum Largest Weighted Delay First (MLWDF) Algorithm If the traffic is a Real Time (RT), the MLWDF is introduced which is used in CDMA-HDR system in order to support RT data users [11].It is more complex algorithms compare with PF and is used in different QoS user s requirements. This is because it takes into account variations of the channel when assigning RBs. Moreover, if a video traffic scenario, it takes into consideration time delay. Any user in MLWDF is granted RBs if it maximizes the equation below: where; (4) (5) where is a difference in time between current and arrival times of the packet that known as the Head Of Line (HOL) packet delay of user i at time t. Similarly to PF equation, while the achievable data rate of user i at time t is, the average data rate of the same user at the same time interval t is. and are the delay threshold for a packet of user i and the maximum HOL packet delay probability of user i respectively. The later is considered to exceed the delay threshold of user i Exponential/Proportional Fair (EXP/PF) Algorithm Since PF is not designed for multimedia applications (only for NRT traffic), an enhanced PF called EXP/PF algorithm was proposed in the Adaptive Modulation and Coding and Time Division Multiplexing (AMC/TDM) systems. The EXP/PF algorithm is designed for NRT service or RT service (different sorts of services). The metric is used for both RT nad Non-RT in which RBs are assigned to users based on. (6) where, (7) (8) where the average number of packets at the buffer of the enb at time t is represented by, k and in equation (8) are constants, is explained in MLWDF, is the HOL packets delay of RT service and is the maximum delay of RT service users. The EXP/PF differentiates between RT and NRT by prioritizing RT traffic users over the NRT traffic users if their HOL values are reaching the delay threshold. 113

129 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October SIMULATION ENVIRONMENT LTE-Sim simulator is used in this paper to do the entire analysis and study [12]. The most recent version of LTE-Sim (version 5) has not involved yet any code regarding the HetNets type (Macro with Pico cells). The developed code used in this paper could be considered as an enhancement of the released LTE-Sim versions. However, LTE-Sim has a detailed code (or what authors are named it: scenario) which can be used to simulate and examine HetNets type (Macro with Femto). Our paper is based on a scenario of a single Macro cell with 2 small Pico cells that are reduced their powers. More Picos can be added to the system, and enhanced system behaviour will be presented. However, according to [2], while the number of Pico cells is increased, more inter-cell interference is experienced since the same carrier frequency is used in each cell (Macro and Picos). Figure 3 shows the entire system that is used in this paper: Macro cell of 1 km and 2 Pico cells of 0.1 km located on the Macro edge. This design is chosen to analog a real system aimed to cover larger area and more users, especially the users in the cell edge where they suffer from lack of connectivity with Macro cell. The inter-cell interference is modeled. Video and VoIP traffic are used to represent user s data. Each user has 50 % Video traffic and 50% VoIP flows. Handover is activated. Each cell starts a certain number of users. Non-uniform user distribution within the cells is applied and 3km/h constant speed is utilized as the mobility user speed. In addition, the 3GPP urban Macro cell propagation loss model has been implemented including path-loss, penetration loss, multi-path loss and shadow fading which are summarized below [13]: Pathloss:, d refers the distance between the enb and the user in kilometers. Penetration loss: 10 db Multipath loss: using one of the well-known methods called Jakes model Shadow fading loss (recently it could be used as a gain in LTE-A): log-normal distribution - Mean value of 0 db. - Standard deviation of 10 db. Figure.3 Applied HetNets (Macro with 2 Picos) 114

130 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October 2014 Packets throughput (see equation 9), Packet Loss Ratio (PLR) as shown in equation 10, packet delay (latency) and fairness index (equation 11) are the concepts used in the aforementioned algorithms to evaluate the system performance. Jain s method is applied to implement fairness among users [14]. According to [1], fairness should reach the value of 1 to be considered as a fair algorithm that sharing the resources suitably among users. It can be calculated as value 1 minus the value of the difference between the maximum and minimum size of transmitted packets of the most and least scheduled users. Equation (11) calculates the fairness value. (11) Obviously, while is the size of transmitted packets, is the size discarded or lost packets during the connection. is the summation of all arrived packets that are buffered into serving enb [1]. The aforementioned total size of transmitted packets of the best served UE and the worse served UE are represented in equation (11) as and. Table 2 shows the entire system simulation parameters [1]. Parameters Simulation time Flow duration Slot duration TTI Number of OFDM symbols/slot Macro cell radius Macro enb Power Pico cell radius Pico enb Power User speed VoIP bit rate Video bit rate Frame structure type Table 2. LTE system simulation parameters 30 s 20 s 0.5 ms 1 ms 7 1 km 49 dbm 0.1 km 30 dbm 3 km/h 8.4 kbps 242 kbps FDD 10 MHz Bandwidth Number of RBs 50 Number of subcarriers 600 Number of subcarriers/rb 12 Subcarrier spacing 15 KHz 115

131 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October 2014 In order to get better results and to confirm the outcomes, five simulations have been conducted for each algorithm (PF, MLWDF and EXP) in each point of users (10, 20, 30, 40, 50, 60, 70 and 80). This yields 120 simulations outcomes. The average values have been taken to draw the simulation graphs at each point of users. 5. SIMULATION RESULTS The average overall system throughput is shown in figure 4. Comparing the throughput for single Macro cell for the same simulation parameters as shown in figure 5, the pico cells in the scenario Macro with 2 Picos boost the throughput by adding gain that shown as an overall system throughput increment for the same number of users. For instance, at 40 users using MLWDF, the throughput is 25 Mbps for the scenario with 2 Picos while the Macro scenario is only 9.3 Mbps. This is almost a duple value. Further points show duple and triple throughput values in the scenario of 2 Picos. However, the gain will reach a saturation level where no more gain could be shown due to the fact of limited radio resources availability while more users are added to the system. Although MLWDF and EXP have almost similar behaviour in both scenarios, a higher throughput is shown in the 2 Pico case using both algorithms. It could note that PF algorithm as shown figure 5 behaves better than the scenario of single Macro cell. PF is developed for NRT traffic, but the simulation is for Video flows (RT traffic); hence, the other simulated algorithms outperform PF. PLR shown in the figure 6 according to [15] is the packet loss ratio for a single Macro cell. While the system is charged with more than 20 users, the PLR is increased for all experienced algorithms taking into consideration that the PF is the worst case with the video traffic. Adding two Picos to the previous system to create Macro with 2 Picos scenario enhances the PLR while maintaining similar system behavior for all algorithms. Approximately, the PLR in Macro with 2 Picos case is reduced to be a quarter of PLR value of single Macro cell scenario. For example, at 70 users, MLWDF has 0.1 PLR value while for the same number of users MLWDF has 0.5 PLR value in the single Macro scenario. Comparing between scheduling schemes, the worst case is the PF algorithm in both cases. Figure 7 illustrates PLR for Macro with 2 Picos. According to [15] and as shown in figure 8, the delay in single Macro cell scenario is close to be constant for PF, MLWDF and EXP/PF with value less than 5 ms while it suffers from rapid increasing after 40 users for PF algorithm. If two Pico cells are added to the aforementioned system, a similar performance is shown, but the delay value is decreased. In addition, the threshold of PF is shifted at 60 users instead of 40 users in the single Macro case. To compare MLWDF and EXP/PF in both scenarios, a certain point in figures 8 and 9 could be explained. For example at 60 users, in a single Macro cell the delay value is 50 ms while in the Macro with 2 Picos the value is 20 ms. As a consequence, for MLWDF and EXP/PF, the delay value with two Picos is approximately half the delay value without Pico cells. One of the purposes of HetNets is to enhance the latency, and this is shown in a practical simulation illustrated in figure 9. However, the delay shows lower values (nearly 10 times lower) in the scenario of single cell with 2 Picos using PF scheme. When the number of users increases in single Macro cell more than 30, the fairness index of all simulated algorithms is deviated down of the value 1. At 40 users, PF shows further deviation close to value 0.8 compare with other algorithms which they are around 0.9. The fairness index behaves similarly in the scenario of Macro with 2 Picos as shown in figure 11. However, the PF shows a minor different in which at 50 users it starts to decline to get the value

132 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October Throughput Figure.4 Average System Throughput (Macro with 2 Picos) Figure.5 Average System Throughput (single Macro cell) 117

133 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October Packet Loss Ratio (PLR) Figure.6 PLR of Video Flows (single Macro cell) [15] Figure.7 PLR of Video Flows (Macro with 2 Picos) 118

134 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October Delay Figure.8 Packet Delay of Video Flows (single Macro cell) Figure.9 Packet Delay of Video Flows (Macro with 2 Picos) 119

135 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October Fairness Index Figure.10 Fairness Index of Video Flows [15] Figure.11 Fairness Index of Video Flows Macro with 2 Picos 120

136 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October CONCLUSION This paper investigates scheduling algorithms that are developed to enhance the LTE network performance by sharing radio resources fairly among users utilizing all available resources. These algorithms depend on traffic class and number of users, hence; different outcomes are presented for each algorithm. To further boost the overall system performance, this study uses heterogeneous networks concept by adding small cells (2 Pico cells). This enhancement is experienced through a throughput, PLR, delay and fairness. In the throughput the system gains more data rate while in PLR the system suffers less packet loss values. Moreover, delay is decreased and fairness stays similar. Approximately from the simulation outcomes, the overall system performance is as follows: throughput is duplicated or nearly tripled relaying on the number of users, the PLR is almost quartered, the delay is reduced 10 times (PF case) and changed to be a half value (MLWDF/EXP cases), and the fairness stays closer to value of 1. As a number of small cells increases, the system is expected to be more enhanced till a saturation state is reached. The reason behind that is the inter-cell interference will limit the performance since the same carrier frequency is used in all system s cells. Focusing on macro with 2 Pico cells scenario, MLWDF shows the best performance for video flows followed by EXP/PF. Further enhancement can be applied in future papers such as almost blank subframes (ABS), enhanced inter-cell interference cancelation (eicic) and cell range extension CRE concepts. REFERENCES [1] H. A. M. Ramli, R. Basukala, K. Sandrasegaran, and R. Patachaianand, "Performance of well known packet scheduling algorithms in the downlink 3GPP LTE system," in Communications (MICC), 2009 IEEE 9th Malaysia International Conference on, 2009, pp [2] Seung June Yi, S.C., Young Dae Lee, Sung Jun Park, Sung Hoon Jung 2012, Radio Protocols for LTE and LTE-Advanced. [3] [1] B. Liu, H. Tian, and L. Xu, "An efficient downlink packet scheduling algorithm for real time traffics in LTE systems," in Consumer Communications and Networking Conference (CCNC), 2013 IEEE, 2013, pp [4] A. Jalali, R. Padovani, and R. Pankaj, "Data throughput of CDMA-HDR a high efficiency-high data rate personal communication wireless system," in Vehicular Technology Conference Proceedings, VTC 2000-Spring Tokyo IEEE 51st, 2000, pp [5] M. Andrews, K. Kumaran, K. Ramanan, A. Stolyar, P. Whiting, and R. Vijayakumar, "Providing quality of service over a shared wireless link," Communications Magazine, IEEE, vol. 39, pp , [6] J.-H. Rhee, J. M. Holtzman, and D. K. Kim, "Performance analysis of the adaptive EXP/PF channel scheduler in an AMC/TDM system," Communications Letters, IEEE, vol. 8, pp , [7] J. Zyren and W. McCoy, "Overview of the 3GPP long term evolution physical layer," Freescale Semiconductor, Inc., white paper, [8] B. Riyaj, M. R. H. Adibah, and S. Kumbesan, "Performance analysis of EXP/PF and M-LWDF in downlink 3GPP LTE system," [9] X. Qiu and K. Chawla, "On the performance of adaptive modulation in cellular systems," Communications, IEEE Transactions on, vol. 47, pp ,

137 International Journal of Wireless & Mobile Networks (IJWMN) Vol. 6, No. 5, October 2014 [10] S. C. Nguyen, K. Sandrasegaran, and F. M. J. Madani, "Modeling and simulation of packet scheduling in the downlink LTE-advanced," in Communications (APCC), th Asia-Pacific Conference on, 2011, pp [11] A. Alfayly, I.-H. Mkwawa, L. Sun, and E. Ifeachor, "QoE-based performance evaluation of scheduling algorithms over LTE," in Globecom Workshops (GC Wkshps), 2012 IEEE, 2012, pp [12] G. Piro, L. A. Grieco, G. Boggia, F. Capozzi, and P. Camarda, "Simulating LTE cellular systems: an open-source framework," Vehicular Technology, IEEE Transactions on, vol. 60, pp , [13] M. Iturralde, T. Ali Yahiya, A. Wei, and A. Beylot, "Resource allocation using shapley value in LTE networks," in Personal Indoor and Mobile Radio Communications (PIMRC), 2011 IEEE 22nd International Symposium on, 2011, pp [14] R. Jain, D.-M. Chiu, and W. R. Hawe, A quantitative measure of fairness and discrimination for resource allocation in shared computer system: Eastern Research Laboratory, Digital Equipment Corporation, [15] AL-Jaradat, Huthaifa 2013, On the Performance of PF, MLWDF and EXP/PF algorithms in LTE. [16] Holma H, Toskala A 2012, LTE-Advanced 3GPP Solution for IMT-Advanced. Authors Haider Al Kim got the B.Sc. in Information and Communication Engineering from Al-khwarizmi Engineering College, University of Baghdad, Baghdad, Iraq in He pursues his Master degree in Telecommunication Networks from University of Technology Sydney (UTS), Sydney, Australia in 2014 under the supervision A. Prof. Kumbesan Sandrasegaran. Working and research areas are Wireless Telecommunication, Mobile Network, Network Management, Network Design and Implementation and Data Analysis and Monitoring.He is senior network engineer with more than 5 years work experience in networks and telecommunication industry at University of Kufa, Iraq. He is also a Cisco Certificate holder (ID: CSCO ) and Cisco instructor at Al-Mansour College, Baghdad, Iraq in Alcatel-Lucent SAM certification holder, Alcatel University, Sydney Australia Shouman Barua is a PhD research scholar at the University of Technology, Sydney. He received his BSc in Electrical and Electronic Engineering from Chittagong University of Engineering and Technology, Bangladesh and MSc in Information and Communication Engineering from Technische Universität Darmstadt (Technical University of Darmstadt), Germany in 2006 and 2014 respectively. He holds also more than five years extensive working experience in telecommunication sector in various roles including network planning and operation. Pantha Ghosal is a Graduate Research Assistant at University of Technology, Sydney. Prior to this, he completed B.Sc in Electrical and Electronic Engineering from Rajshahi University of Engineering & Technology, Bangladesh in He is an expert of Telecommunication network design and holds more than 7 years of working experience in this area. Dr Kumbesan Sandrasegaran is an Associate Professor at UTS and Centre for Real-Time Information Networks (CRIN). He holds a PhD in Electrical Engineering from McGill University (Canada)(1994), a Master of Science Degree in Telecommunication Engineering from Essex University (1988) and a Bachelor of Science (Honours) Degree in Electrical Engineering (First Class) (1985). His current research work focuses on two main areas (a) radio resource management in mobile networks, (b) engineering of remote monitoring systems for novel applications with industry through the use of embedded systems, sensors and communications systems. He has published over 100 refereed publications and 20 consultancy reports spanning telecommunication and computing systems. 122