在無線網路中針對優先權傳送之排程技術研究

國

國

國

國 立

立

立 交

立

交

交 通

交

通

通 大

通

大

大 學

大

學

學

學

資訊工程學系

博士論文

博士論文

博士論文

博士論文

在無線網路中針對優先權傳送之

在無線網路中針對優先權傳送之

在無線網路中針對優先權傳送之

在無線網路中針對優先權傳送之

排程

排程

排程

排程技術研究

技術研究

技術研究

技術研究

Scheduling Techniques for Priority

Transmission in Wireless Networks

研究生：賈仲雍

指導教授：張明峰 博士

中華民國九十八年

中華民國九十八年

中華民國九十八年

中華民國九十八年七

七

七

七月

月

月

月

在無線網路中針對優先權傳送之排程技術

在無線網路中針對優先權傳送之排程技術

在無線網路中針對優先權傳送之排程技術

在無線網路中針對優先權傳送之排程技術研究

研究

研究

研究

Scheduling Techniques for Priority Transmission in

Wireless Networks

研 究 生：賈仲雍 Student： Chung-Yung Chia

指導教授：張明峰 博士 Advisor： Dr. Ming-Feng Chang

國 立 交 通 大 學

資 訊 工 程 學 系

博 士 論 文

A Dissertation

Submitted to Institute of Computer Science and Engineering College of Computer Science

National Chiao Tung University in partial Fulfillment of the Requirements

for the Degree of Doctor of Philosophy

in

Computer Science and Engineering July 2009

Hsinchu, Taiwan, Republic of China

i

在無線網路中針對優先權傳送之排程技術

在無線網路中針對優先權傳送之排程技術

在無線網路中針對優先權傳送之排程技術

在無線網路中針對優先權傳送之排程技術研究

研究

研究

研究

學生：賈仲雍

指導教授：張明峰 博士

國立交通大學資訊科學與工程研究所

摘要

摘要

摘要

摘要

在越來越競爭的電信市場上，電信業者如何能同時滿足用戶服務的滿意度和保持各 項業務的盈收是相形重要。論文中針對幾種不同無線網路型態(如 GPRS, 3G , HSDPA) 的用戶在使用無線資源時，除考量傳送或接收封包優先權排程方式之外，另針對吃到飽 ( flat-rate)用戶，考量其對行動業者的最有效率優先權排程機制，做了廣泛且深入之研究。 首先在 GSM 行動網路中，我們研究和比較 3 種不同排程方式，如何在用戶傳送上 傳(uplink)封包時，控制和調整用戶的上傳封包的傳送優先權，透過建立數學分析模式和 幾種不同分析參數，比較何種優先權考量可滿足用戶服務滿意度。研究發現利用上鏈路 狀態旗標(USF)配置的排程方式可以讓有較高優先權封包(如 VoIP)，其傳輸延遲可以比 較少。 接著，在 UMTS 行動網路中，我們研究和比較 4 種不同排程方式，如何在用戶啟 動上傳連線(uplink connection transmission)時，控制調整用戶的上傳連線的傳送優先權， 透過建立數學分析模式、幾種不同分析參數和考量吃到飽用戶所提出的代價函數(cost

function)，來比較何種優先權考量可同時滿足用戶服務滿意度和保持電信業者盈收。研

究發現單憑降低用戶的上傳連線速度，並不會讓整體用戶的上傳連線傳送效能最高和讓 電信業者擁有最大盈收，若利用等待隊列(waiting queue)和強制取代(preemption)的排程 方式，可以同時兼顧用戶服務滿意度和讓電信業者擁有最大盈收。

ii 最後，在 HSDPA 行動網路中，我們研究和比較 4 種不同排程方式，如何在用戶傳 送下傳(downlink)封包時，控制調整用戶的下傳封包的傳送優先權，透過建立數學分析 模式、幾種不同分析參數和考量吃到飽用戶所提出的代價函數(cost function)，比較何種 優先權考量可同時滿足用戶服務滿意度和保持電信業者盈收。研究發現若利用等待隊列 (waiting queue)和固定強制取代(preemption)的排程方式，並不會讓整體用戶的下傳封包 傳送效能最高和讓電信業者擁有最大盈收，若考量加入動態丟棄計時器(Drop Timer)和 動態防護通道(Guard Slot)的排程方式，來動態調整用戶下傳封包的傳送優先權，可以同 時兼顧用戶服務滿意度和讓電信業者擁有最大盈收。 在無線網路發展趨勢上，電信業者必須持續不斷的研究如何能同時滿足用戶服務的 滿意度和保持各項業務的盈收。本篇論文研究的結果可被當成繼續研究在無線網路中如 何考量優先權排程方式的基礎。。。。 關鍵字 關鍵字 關鍵字 關鍵字：：：優先權傳送、點陣圖通道配置、上鏈路狀態旗標、吃到飽服務、連線排程、動： 態丟棄計時器和動態防護通道。

iii

Scheduling Techniques for Priority Transmission in Wireless

Networks

Student: Chung-Yung Chia Advisor: Dr. Ming-Feng Chang

Institute of Computer Science and Engineering

National Chiao-Tung University

Abstract

In today’s highly competitive telecommunications environment, the emphasis has shifted

to delivering innovative services to satisfy increasingly sophisticated customers’ need and to

improve the revenues of wireless operators. This dissertation has studied different scheduling

mechanisms for priority transmission in public land mobile networks (PLMNs). We

considered not only the priority in packet transmission/reception using various scheduling

techniques, but also the most efficient mechanisms for serving both normal and flat-rate

customers.

First, we study and compare three different scheduling mechanisms in the GSM network.

As the General Packet Radio Service (GPRS) network begins to provide such as

"push-to-talk" (PTT) service, delay-sensitive packets should be given higher priority in

iv

queues for priority packets in the GPRS network: Bitmap Channel Allocation (BCA) and

Uplink State Flag Channel Allocation (USFCA). Our study shows that the transmission delay

of priority packets in the GPRS network can be better guaranteed using USFCA.

Second, we study and compare four different scheduling mechanisms in the UMTS

network. To attract more users to mobile packet services, the Universal Mobile

Telecommunication System (UMTS) operators have been prompting flat-rate packet services.

Since usage does not incur cost, flat-rate users tend to stay on line longer and occupy most of

the radio channel resources. We consider a UMTS network serving two types of user

connections: Normal User Connections (NUCs) and Flat-Rate User Connections (FRUCs).

Our goal is to maximize the revenue of the operator by giving a priority to NUCs over FRUCs

without discontenting the flat-rate users, in order not to lose the flat-rate users to other

operators. Uplink FRUCs may be asked to sub-rate or suspend transmission when the radio

network is fully utilized. Four combinations of scheduling techniques including queueing,

guard channels, preemption and rate-adaptation, have been studied, and analytic models using

Markov processes were used to evaluate their performances. We proposed a cost function

representing the revenue loss due to both blocked NUCs and lost flat-rate users. The system

parameters used in our analysis are based on realistic operation data. Our analytic results

indicate that the revenue loss can be minimized by using waiting queues and preemption.

Rate-adaptation is ineffective in minimizing the revenue loss because sub-rated connections

are less efficient in using radio spectrum. Guard channels for NUCs are unnecessary when

waiting queue or preemption is used. Our study may be valuable for UMTS operators in

serving flat-rate users.

Third, we study and compare four different scheduling mechanisms in the HSDPA

network. We consider a HSDPA network serving two types of user packets: charged packets

(CPs) and flat-rate packets (FRPs). Since CPs are charged by usage, they are given a higher

v

preference may lead to poor quality of service for flat rate users. In particular, FRPs may

experience longer transmission latency and higher dropped probability. We should consider

the balance between serving the FRPs and CPs. Analytic models using Markov process were

used to study their performance. Our study shows that DDT-PQ and DGS-PQ methods are

more effective to transmit the downlink CPs especially when the downlink FRP traffic is high.

Therefore, they are better in guaranteeing the system throughput for CPs, and thus the

operator revenue can be better protected.

In the development trend of wireless network, wireless operators need to keep studying

how to satisfy the QoS requirements of the customers and have the best revenues at the same

time. The research results presented in this dissertation can be viewed as a useful foundation

for further study in the scheduling mechanisms for priority transmission in the wireless

network.

Key Words: Priority Transmission, Bitmap Channel Allocation (BCA), Uplink State Flag

(USF), Flat-Rate Service, Connection Scheduling, Dynamic Discard Timer (DDT) and

vi

Acknowledgement

I would like to express my sincere thanks to my advisors, Prof. Ming-Feng Chang.

Without their supervision and perspicacious advice, I can not complete this dissertation.

Special thanks to my committee members, Prof. Chien-Chao Tseng, Prof. Kuo-Chen Wang,

Prof. Duan-Shin Lee, Prof. Ai-Chun Pang, Dr. Sheng-Lin Chou and Dr. Kuang-Yao Chang for

their valuable comments. Thanks also to the colleagues in Internet Communication

Laboratory.

I also express my appreciation to all the faculty, staff and colleagues in the Institute of

Computer Science and Engineering, NCTU and Wireless Communications Laboratory,

Telecommunication Laboratories of CHT. In particular, I would like to thank Dr. Chen and Dr.

Yoau for their friendship and support in various ways.

Finally, I am grateful to my family, my wife, Shiely, my children Arron and Sarah, and

vii

Contents

Abstract in Chinese... i

Abstract in English ...iii

Acknowledgement ... vi

Contents ... vii

List of Figures... x

List of Tables ...xii

Abbreviation ...xiiiii

CHAPTER 1 Introduction ... 1

1.1 Channel Allocation for Priority Packets in the GPRS Network ... 1

1.2 Uplink Connection Scheduling for Flat-Rate Data Services in the UMTS Network ... 2

1.3 Flate-Rate Packet Scheduling for the WCDMA Systems with HSDPA ... 2

1.4 Synopsis of This Dissertation ... 3

CHAPTER 2 Channel Allocation for Priority Packets in the GPRS Network... 4

2.1 Introduction ... 4

2.2 The Methods of BCA and USFCA ... 4

2.2.1 BCA Method... 5

2.2.2 USFCA Method ... 5

2.3 The Analytic Models ... 6

2.3.1 BCA Method... 7

2.3.2 USFCA Method ... 9

2.4 Numeric Results ... 10

viii

CHAPTER 3 Uplink Connection Scheduling for Flat-Rate Data Services

in the UMTS Network ... 13

3.1 Introduction ... 13

3.2 System Models and Assumptions... 19

3.2.1 CDMA Uplink Capacity Model...22

3.2.2 A Scheduler with all four features...23

3.3 The Analytic Models ... 25

3.3.1 The Analytic Model of SAll...25

3.3.2 The Performance Measures...28

3.3.3 Cost Function Scheme...32

3.3.4 An Iterative Algorithm ...34

3.4 Numerical results and Discussions ... 35

3.5 Conclusions ... 44

CHAPTER 4 Flate-Rate Packet Scheduling for the WCDMA Systems with HSDPA ...46

4.1 Introduction ... 46

4.2 HSDPA Basic Principles... 48

4.3 System Models and Assumptions ... 51

4.3.1 M-PQ Method... 52

4.3.2 P-PQ Method ... 52

4.3.3 DDT-PQ Method...54

4.3.4 DGS-PQ Method...56

4.4 The Analytic Method ... 59

4.4.1 M-PQ Method... 60

4.4.2 P-PQ Method ... 61

ix

4.4.4 DGS-PQ Method...64

4.5 Cost Function Scheme ... 66

4.6 Numreical Analysis...68

4.7 Conclusions ... 77

CHAPTER 5 Conclusions and Future Work... 79

5.1 Summary... 79

5.2 Future Works ... 81

Bibliography ...83

Curriculum Vitae ... 88

x

List of Figures

Fig. 2.1: A USFCA example using priority-packet-first scheme ...6

Fig. 2.2: The queuing model for BCA and USFCA schemes ...7

Fig. 2.3: The mean waiting time and system time of uplink packets ...11

Fig. 3.1: The Radio Access Bearer (RAB) assignment procedure ...20

Fig. 3.2: The system queueing model for a reference cell ...24

Fig. 3.3: An algorithm for the numbers of full-rate and half-rate FRUCs in state (i, j,k)...26

Fig. 3.4: The state transition diagram of S_All, and the rates of input/output flows...27

Fig. 3.5: An iterative algorithm minimizes the cost function...34

Fig. 3.6.a:The cost function (C) with

α

=0.504 and



=0.02, (B=4, Q=10)...36

Fig. 3.6.b:The numbers of guard channels (GC) with

α

=0.504 and



=0.02, (B=4, Q=10).37 Fig. 3.7.a: Average NUC blocking probabilities (PBN) with B=4, Q=10...38

Fig. 3.7.b: Average FRUC blocking probabilities (PBF) with B=4, Q=10...39

Fig. 3.8.a: Average NUC waiting times (WTN) with B=4, Q=10...39

Fig. 3.8.b: Average FRUC waiting times (WTF) with B=4, Q=10...40

Fig. 3.9.a: Average NUC queueing probabilities (PQN) with B=4, Q=10...41

Fig. 3.9.b: Average FRUC queueing probabilities (PQF) with B=4, Q=10...41

Fig. 3.10.a: Average preempted probabilities of a serving full-rate FRUCs (PFPrm) with B=4, Q=10...42

Fig. 3.10.b: Average preempted probabilities of a serving half-rate FRUCs ( PSPrm) with B=4, Q=10...42

Fig. 3.11:Average sub-rated probabilities of a serving full-rate FRUCs (PFS) (B=4, Q=10) 43 Fig. 3.12:Average transmission rate of serving FRUCs (TF) (B=4, Q=10) ...44

xi

Fig. 4.2: Downlink SF codes allocation tree for HS-DSCH and HS-SCCH...50

Fig. 4.3: An example downlink packet scheduling for MS1-MS4 in a cell...50

Fig. 4.4: The scheduling parameters sent from an SGSN through a RNC to a Node-B in HSDPA network...51

Fig. 4.5: A queueing model for M-PQ scheme ...53

Fig. 4.6: A queueing model for P-PQ scheme ...53

Fig. 4.7: A queueing model for DDT-PQ scheme ...54

Fig. 4.8: A peudocode for DDT-PQ scheme ...56

Fig. 4.9: A queueing model for DGS-PQ scheme ...57

Fig. 4.10: A peudocode for DGS-PQ scheme ...59

Fig. 4.11: The state transition diagram of M-PQ scheme ...60

Fig. 4.12: The state transition diagram of P-PQ scheme ...62

Fig. 4.13: The state transition diagram of DDT-PQ scheme ...63

Fig. 4.14: The state transition diagram of DGS-PQ scheme ...65

Fig. 4.15: The average dropped probability of CP for four packet scheduling methods ...70

Fig. 4.16: The average dropped probability of FRP for four packet scheduling methods ...71

Fig. 4.17: The average network utilization of CP for four packet scheduling methods ...72

Fig. 4.18: The results of dynamically adjusting the DT value in DDT-PQ scheme ...73

Fig. 4.19: The results of dynamically adjusting the GS value in DGS-PQ scheme ...74

Fig. 4.20: The cost function (C) when

α

>ρ...75

Fig. 4.21: The number of GS for CPs and FRPs when

α

>ρ...76

Fig. 4.22: The value of DT for FRPs and CPs when

α

>ρ...76

xii

List of Tables

xiii

Abbreviation

The abbreviations used in this dissertation are listed below.

3GPP: 3rd Generation Partnership Project

AP: Access Point

BCA:Bitmap Channel Allocation

BSS: Business Support System

CN:Core Network

CS: Circuited-Switch

CP: Charged Packets

CPICH: Common Pilot Indicator Channel

D_CH: Dedicated radio Channel

DDT: Dynamic Discard Timer

DGS: Dynamic Guard Slots

FDD:Frequency Division Duplex

FCFS:First Come, First Served

FRP: Flat-Rate Packets

FRUC:Flat-Rate User Connections

GC: Guard Channels

GGSN: Gateway GPRS Support Node

GPRS: General Packet Radio Service

GSM: Global System for Mobile Communication

HSDPA: High Speed Downlink Packet Access

xiv HS-SCCH : High Speed Shared Control Channel

LB: Lower Bound

Max C/I: Maximum Carrier-to-Interference ratio

MCS: Modulation Coding Scheme

MS: Mobile Station

NUC:Normal User Connections

PASTA:Poisson Arrivals See Time Averages

PDP: Packet Data Protocol

PF: Proportionally Fair

PLMN: Public Land Mobile Network

PS: Packet-Switched

PSTN: Public Switched Telephone Network

PTT: Push To Talk

PQ: Priority Queue

QAM: Quadrature Amplitude Modulation

QoS:Quality of Service

QPSK: Quadrature Phase-Shift Keying

RAB: Radio Access Bearer

RAU:Routing Area Upate

RB:Radio Bearer

RF: Radio Frequency

RNC:Radio Network Controller

RR: Round Robin

RRC:Radio Resource Control

SF: Spreading Factor

xv SIR: Signal to Interference Ratio

SV: Step Value

TDMA:Time Division Multiple Access

TTI: Transmission Time Interval

UB: Upper Bound

UMTS: Universal Mobile Telecommunications System

URI: Universal Resource Identifier

USFCA:Uplink State Flag Channel Allocation

VoIP: Voice over IP

UTRAN:UMTS Terrestrial Radio Access Network

WCDMA: Wideband CDMA

WLAN: Wireless Local Area Network

1

CHAPTER 1

Introduction

In recent years, telecommunication industry is growing fast especially in mobile market.

Many technologies have been developed and deployed, such as 2G/GPRS/3G/HSDPA, VoIP,

and NGN (Next Generation Network). They provide not only the traditional voice

communication service but also many advanced data and information services. However,

technical advances no longer drive the market trend. In today’s highly competitive

environment, the emphasis has shifted to delivering innovative service to satisfy increasingly

sophisticated customers’ need. Customers have changed from being the passive role to active,

and operators need to focus on customers’ feeling. It is important to satisfy all kinds of

customers and make them feel that the service is tailored for them, for their benefits and

interests. Of coourse, wireless operators also need to consider the revenues when they provide

the services to the customers. To consider the customer’s need and the revenues of the

operators, they need effective scheduling mechanisms to process the customer

uplink/downlink packets. Operators shall make a win-win new telecom business market.

1.1 Channel Allocation for Priority Packets in the GPRS Network

As the General Packet Radio Service (GPRS) network begins to provide such as

"push-to-talk" (PTT) service, delay-sensitive packets should be given higher priority in

transmission. In this chapter, we study two channel allocation schemes that implement

2

and Uplink State Flag Channel Allocation (USFCA). Our study shows that the transmission

delay of priority packets in the GPRS network can be better guaranteed using USFCA.

1.2 Uplink Connection Scheduling for Flat-Rate Data Services in the

UMTS Network

To attract more users to mobile packet services, the Universal Mobile

Telecommunication System (UMTS) operators have been prompting flat-rate packet services.

Since usage does not incur cost, flat-rate users tend to stay on line longer and occupy most of

the radio channel resources. We consider a UMTS network serving two types of user

connections: Normal User Connections (NUCs) and Flat-Rate User Connections (FRUCs).

Our goal is to maximize the revenue of the operator by giving a priority to NUCs over FRUCs

without discontenting the flat-rate users, in order not to lose the flat-rate users to other

operators. Uplink FRUCs may be asked to sub-rate or suspend transmission when the radio

network is fully utilized. Four combinations of scheduling techniques including queueing,

guard channels, preemption and rate-adaptation, have been studied, and analytic models using

Markov processes were used to evaluate their performances. We proposed a cost function

representing the revenue loss due to both blocked NUCs and lost flat-rate users. The system

parameters used in our analysis are based on realistic operation data. Our analytic results

indicate that the revenue loss can be minimized by using waiting queues and preemption.

Rate-adaptation is ineffective in minimizing the revenue loss because sub-rated connections

are less efficient in using radio spectrum. Guard channels for NUCs are unnecessary when

waiting queue or preemption is used. Our study may be valuable for UMTS operators in

serving flat-rate users.

1.3 Flate-Rate Packet Scheduling for the WCDMA Systems with HSDPA

3

provide flat-rate packet services in the WCDMA system with High Speed Downlink Packet

Access (HSDPA). Since usage does not incur extra cost, flat-rate users may always stay on

line and occupy most of the radio channel resources. In this chapter, we consider a HSDPA

network serving two types of user packets: charged packets (CPs) and flat-rate packets (FRPs).

Since CPs are charged by usage, they are given a higher priority to receive downlink packets

for revenue consideration. However, this priority preference may lead to poor quality of

service for flat rate users. In particular, FRPs may experience longer transmission latency and

higher dropped probability. We should consider the balance between serving the FRPs and

CPs. Four downlink packet scheduling methods are studied in this chapter: (1) Max. C/I first

in a priority queue (M-PQ) ; (2) CPs first in a PQ (P-PQ) ; (3) Dynamic discard timer for

FRPs in a PQ (DDT-PQ) and (4) Dynamic guard slots for CPs in a PQ (DGS-PQ). Analytic

models using Markov process were used to study their performance. Our study shows that

DDT-PQ and DGS-PQ methods are more effective to transmit the downlink CPs especially

when the downlink FRP traffic is high. Therefore, they are better in guaranteeing the system

throughput for CPs, and thus the operator revenue can be better protected.

1.4 Synopsis of This Dissertation

This dissertation is organized as follows. Chapter 2 presents Channel Allocation for

Priority Packets in the GPRS Network. Chapter 3 presents Uplink Connection Scheduling for

Flat-Rate Data Services in the UMTS Network. Chapter 4 presents Flate-Rate Packet

Scheduling for the WCDMA Systems with HSDPA. Chapter 5 concludes this dissertation and

4

CHAPTER 2

Channel Allocation for Priority Packets in the GPRS

Network

2.1 Introduction

General Packet Radio Service (GPRS) has been developed to provide packet data

services based on the circuit-switching GSM network. Much research has been done on

analyzing the performance of fixed or dynamic channel (i.e., timeslots) allocation to support

multiple-slot data transmission [1-2]. However, very few studies considered special

treatments to priority packets in the GPRS network. In [3], Chew and Tafazolli give priority

to mobility management packets to ensure minimal delay. Their results indicated that the

priority queue provides shorter RAU completion time and higher packet throughput than the

others. However, the way in which the priority queue is implemented in the GPRS network

has not been thoroughly studied.

In addition to the mobility management packets, some data services, such as "push to

talk" (PTT) are delay-sensitive; the transmission latency of voice packets is very important to

the quality of the communications. In this chapter, we study two channel allocation schemes

[4], Bitmap Channel Allocation (BCA) and Uplink State Flag Channel Allocation (USFCA),

that implement priority queues to give transmission priority to packets requiring shorter

transmission latency. We also present analytic models to analyze their performance in terms

of packet transmission delay.

5

A GSM/GPRS TDMA frame consists of eight timeslots, numbered 0-7, which can be

used for data or voice transmission. Channel allocation in the GPRS network can be

performed in unit of radio blocks. A radio block consists of four identical timeslots from four

successive TDMA frames. Uplink packet requests from a mobile station (MS) can specify

different priorities for special treatment by the GPRS network [4]. In this chapter, we assume

only two types of packets: priority packets that are sensitive to delay, and non-priority packets

that are not.

2.2.1 BCA Method

For an uplink “Packet Channel Request” message from a MS, the GPRS network

may return a “Packet Uplink Assignment” message with the allocation_bitmap element

indicating the allocated radio blocks to the uplink packet request. To reduce the number

of messages exchanged between the MS and the network, the network allocates radio

blocks in full amount requested by the MS. As a result, when all timeslots of the network

are assigned out, new uplink packet requests need to wait until a transmitting packet

completes. The transmitting packets cannot be interrupted during transmission.

2.2.2 USFCA Method

For an uplink “Packet Channel Request” message from a MS, the GPRS network

may return a “Packet Uplink Assignment” message with a

USF_for_each_timeslot_number element indicating a specific USF value for each

timeslot allocated to the uplink packet request. For USFCA, the network broadcasts a

USF value at each downlink radio block. In the next uplink radio block, the MS assigned

with the same USF value can transmit for one radio block. In this way, the network can

schedule an uplink packet to transmit at the next radio block on a radio block by radio

6

block. The way in which multiple packets shares a timeslot is controlled the network; the

network can use various scheduling schemes, such as priority-packet-first. Fig. 2.1

shows a USFCA example using priority-packet-first scheme, a non-priority packet 1 is

assigned with USF value 1 and a priority packet 2, which needs m radio blocks to

transmit, is assigned with USF value 2 by network. Packet 1 is transmitting when packet

2 arrives at radio block n. The network suspends the transmission of packet 1, and

instructs packet 2 to transmit at downlink radio block n+1. Packet 1 can resume

transmission after packet 2 completes transmission.

Fig. 2.1: A USFCA example using priority-packet-first scheme

2.3 The Analytic Models

Let C denote the number of GPRS timeslots reserved for transmission of data packets.

When all the GPRS timeslots are assigned, additional uplink packet requests are put in a

priority queue of size B maintained by the network. In the priority queue, packets of the same

priority will be served on a FCFS basis. The queuing model of BCA and USFCA schemes is

depicted in Fig. 2.2. Using BCA, the network cannot suspend the transmission of a packet

under service, but using USFCA, the network can suspend the transmission of a non-priority

uplink radio blocks for transmitting Packet 1

(USF=1, non-priority packet)

-uplink radio blocks for transmitting Packet 2

(USF=2, priority packet) Downlink radio block -Uplink radio block radio block n

Transmit packet 1 Transmit packet 2

radio block n+1 radio block n+m radio block n+m+1

Resume to transmit packet 1

…

USF=2

USF=1

…

USF=2 USF=1

…

radio block n+1 radio block n+2 radio block n+m+1 radio block n+m+2

…

…

…

…

…

uplink radio blocks for transmitting Packet 1

(USF=1, non-priority packet)

-uplink radio blocks for transmitting Packet 2

(USF=2, priority packet) Downlink radio block -Uplink radio block radio block n

Transmit packet 1 Transmit packet 2

radio block n+1 radio block n+m radio block n+m+1

Resume to transmit packet 1

…

USF=2

USF=1

…

USF=2 USF=1

…

radio block n+1 radio block n+2 radio block n+m+1 radio block n+m+2

…

…

…

…

…

7

packet, put it back to the priority queue, and start transmitting a new priority packet. This difference is depicted in Fig. 2.2 by dotted line e.

To analyze the performance of the schemes, we made the following assumptions. The arrivals of priority and non-priority packets form Poisson processes with mean

λ

P and

λ

nP respectively. The service time of priority and non-priority packets is assumed to be

exponentially distributed with mean 1

µ

_P and 1

µ

_nP respectively. We can use the M/M/C/B Markov process to model BCA and USFCA.

Fig. 2.2: The queuing model for BCA and USFCA schemes

2.3.1 BCA Method

In this process, state (i,j,k) denotes that there are i priority packets transmitting in the network, j priority packets waiting in the priority queue, k non-priority packets transmitting in the network or waiting in the priority queue. Let Pi,j,k denote the

steady-state probability of the network in state (i,j,k) and Sbm be the set of existing

states for this process.

(2.1) Priority queue of size B Serving timeslots e. Suspend to the USF scheme) a. A new uplink request R_New b. R_Newis rejected if the queue is full

c. Start to transmit . . … C . . 1 2 d. Transmission completes Priority queue of size B Serving timeslots e. Suspend to the USF scheme) a. A new uplink request R_New b. R_Newis rejected if the queue is full

c. Start to transmit . . . . … C . . 1 2 d. Transmission completes ≤ ≤ + ≤ + + ≤ ={(i,j,k)|0 i j k (C B),0 i C,0 bm

S

bm={(i,j,k)|0≤i+ j +k ≤(C+ B),0≤i≤C,0

S

} = ≥ + + ≤ ≤ ≤ ≤ j≤B,0≤k≤(C+B)and((i+k)≥Cor (j=0))} ≤ j B,0 k (C B)and((i k) Cor (j 0)) ≤ ≤ + ≤ + + ≤ ={(i,j,k)|0 i j k (C B),0 i C,0 bm

S

bm={(i,j,k)|0≤i+ j +k ≤(C+ B),0≤i≤C,0

S

} = ≥ + + ≤ ≤ ≤ ≤ j≤B,0≤k≤(C+B)and((i+k)≥Cor (j=0))} ≤ j B,0 k (C B)and((i k) Cor (j 0))

8

To handle the non-existing states, an indicator θi ,,jk is used to indicate whether state

(i,j,k) exists or not, i.e.,θi ,,jk=1 if state (i,j,k) belongs toSbm. In addition, δ1-δ6 indicators

are used to indicate whether a specific transition exists or not. The balance equations for this process can be expressed in (2.2) and the parameters are defined in (2.3)-(2.10).

(2.2) (2.3) (2.4)

δ

1=1, if (i+k) < C; 0, otherwise. (2.5)

δ

2=1, if (i+k)≥C; 0, otherwise. (2.6)

δ

3=1, if (j==0); 0, otherwise. (2.7)

δ

4=1, if ((i+k)≥C) and (i≠0); 0, otherwise. (2.8)

δ

5=1, if (j==0) and (i≠C); 0, otherwise. (2.9)

δ

6=1, if (i+j+k)≤C; 0, otherwise. (2.10)

From the balance equations and the constraints = 1, the steady-state probabilityPi,j,k can be obtained by an iterative algorithm [5]. The blocking probability

of packets (P_bm ) ; the mean waiting time and system time of priority packets ( Wp_bm and

Tp _bm _{) ; the mean waiting time and system time of non-priority packets (}Wnp_bm and Tnp _bm)

can be expressed in (2.11)-(2.15).

(

)

  _{+ +} − − + + +

δ

λ

θ

θ

θ

λ

δ

1 1,, 2 , 1, 1, 1, 1 , ,jk p i jk p ij k np(, , ) i j k i

M

i jk

P

θ

λ

δ

θ

δ

θ

δ

4

P

i,j+1,k

M

p(i,j+1,k) i,j+1,k+ 5

P

i,j,k+1

M

np(i,j,k+1) i,j,k+1+ 6

P

i−1,j,k p i−1,j,k +   + + + +

_δ

3

M

p(i,j,k)

θ

i₋1,j,k

_δ

4

M

p(i,j,k)

θ

i,j₋1,k

_δ

5

M

np(i,j,k)

θ

i,j,k₋1

λ

np

θ

i,j,k₊1

(

λ

θ

θ

)

δ

θ

δ

2

P

i,j−1,k p i,j−1,k

+

P

i−1,j+1,k+1

M

np(i−1,j+1,k+1) i−1,j+1,k+1

+

3

P

_i+1,_j,_k

M

p(i+1,j,k) i+1,j,k

=

θ

λ

,, 1 1 , , − − +

_P

_{ijk np ijk}

(

)

  _{+ +} − − + + +

δ

λ

θ

θ

θ

λ

δ

1 1,, 2 , 1, 1, 1, 1 , ,jk p i jk p ij k np(, , ) i j k i

M

i jk

P

θ

λ

δ

θ

δ

θ

δ

4

P

i,j+1,k

M

p(i,j+1,k) i,j+1,k+ 5

P

i,j,k+1

M

np(i,j,k+1) i,j,k+1+ 6

P

i−1,j,k p i−1,j,k +   + + + +

_δ

3

M

p(i,j,k)

θ

i₋1,j,k

_δ

4

M

p(i,j,k)

θ

i,j₋1,k

_δ

5

M

np(i,j,k)

θ

i,j,k₋1

λ

np

θ

i,j,k₊1

(

λ

θ

θ

)

δ

θ

δ

2

P

i,j−1,k p i,j−1,k

+

P

i−1,j+1,k+1

M

np(i−1,j+1,k+1) i−1,j+1,k+1

+

3

P

_i+1,_j,_k

M

p(i+1,j,k) i+1,j,k

=

θ

λ

,, 1 1 , , − − +

_P

_{ijk np ijk} l m,n l p p( , ) =

µ

M

l m,n l p p( , ) =

µ

M

l m,n l p p( , ) =

µ

M

l m,n l p p( , ) =

µ

M

otherwise n m l n np np np M ( , , ) µ , µ (C -l) , if ( n (C -l)) n , ) , = ≥ otherwise n m l n np np np M ( , , ) µ , µ (C -l) , if ( n (C -l)) n , ) , = ≥ otherwise n m l n np np np M ( , , ) µ , µ (C -l) , if ( n (C -l)) n , ) , = ≥ otherwise n m l n np np np M ( , , ) µ , µ (C -l) , if ( n (C -l)) n , ) , = ≥ otherwise n m l n np np np M ( , , ) µ , µ (C -l) , if ( n (C -l)) n , ) , = ≥ otherwise n m l n np np np M ( , , ) µ , µ (C -l) , if ( n (C -l)) n , ) , = ≥ ∑ ∈Sbm k j i k j i P ) , , ( , ,

9 (+ +

∑

= + ) = B C k j i k , j , i bm _ P

P

(2.11)

(

)

(

∑

)∈ ⋅ ⋅ − = bm S k , j , i k , j , i bm _ p bm _ p j P

P

W

1 1 λ (2.12)

(

)

(

∑

)∈

(

)

⋅ + ⋅ − = bm S k , j , i k , j , i bm _ p bm _ p i j P

P

T

1 1 λ (2.13)

(

)

( )∈

∑

[

>( )−

(

)

]

⋅ − − ⋅ − = i c k , S k , j , i k , j , i bm _ np bm _ np bm P i C k

P

W

1 1 λ (2.14)

(

)

(

∑

)∈ ⋅ ⋅ − = bm S k , j , i k , j , i bm _ np bm _ np k P

P

T

1 1 λ (2.15)

2.3.2 USFCA Method

In this process, state (i,j) denotes that there are i priority packets and j non-priority packets transmitting in the network or in the priority queue. Let Pi ,j denote the

steady-state probability of the network in state (i,j) and SUSFbe the set of existing states

for this process. x

} + ≤ ≤ + ≤ ≤ + ≤ + ≤ = i,j i j C B i C B j C B SUSF {( ) |0 ,0 ,and0 (2.16)

To handle the un-eisting states, an indicator θi,j is used to indicate whether state (i,j)

exists or not, i.e., θi,j=1 if state (i,j) belongs toSUSF.The balance equations for this

process can be expressed in (2.17) and the parameters are defined in (2.18)-(2.19).

(2.17)

(2.18)

(2.19)

From the balance equations and the constraints(i,j∑)∈SUSFPi,j = 1, the steady-state

probabilityPi ,j can be derived by an iterative algorithm. The blocking probability of

= otherwise m C n m p p p , ) , (

µ

µ

M ≥ m if( C) , = otherwise m C n m p p p , ) , (

µ

µ

M ≥ m if( C) , ≥ + − ≥ otherwise n n m if m m if n m np np np M , C) ) (( , ) C ( ) C ( , 0 ) , (

µ

µ

= − + ≥ ≥ otherwise n n m if m m if n m np np np M , C) ) (( , ) C ( ) C ( , 0 ) , (

µ

µ

=

(

)

θ

θ

θ

λ

θ

λ

θ

λ

θ

θ

θ

λ

1 , 1 , , 1 , 1 1 , 1 , , 1 , 1 1 , , 1 1 , , 1 , ) 1 , ( ) , 1 ( ) , ( ) , ( + + + + − − − − − − + + + + + + + = + + + j i np j i j i p j i j i np j i j i p j i j i np j i p j i np j i p j i j i j i j i j i

M

P

M

P

P

P

M

M

P

10

packets (P_USF) ; the mean waiting time and system time of priority packets ( W p _USF and

Tp _USF) ; the mean waiting time and system time of non-priority packets ( Wnp _USF and

Tnp _USF) can be expressed in (2.20)-(2.24).

(+

∑

= + ) = B C j i j , i USF _ P

P

(2.20)

(

)

( )∈

∑

(

(> )

)

⋅ − ⋅ − = C i , S j , i j , i USF _ p USF _ p USF P C i

P

W

1 1 λ (2.21)

(

)

( )

∑

∈ ⋅ ⋅ − = USF S j , i j , i USF _ p USF _ p i P T

P

1 1 λ (2.22)

(

)

( ) ( ) ( ) (

(

) ( )

)

     ⋅ − + + ⋅ ⋅ − =

∑

∑

> + < ∈ ≥ ∈ i,j S ,i cC ,i j C j , i C i , S j , i j , i USF _ np USF _ np USF USF P C j i P j

P

W

1 1 λ (2.23)

(

)

( )

∑

∈ ⋅ ⋅ − = USF S j , i j , i USF _ np USF _ np j P T

P

1 1 λ (2.24)

2.4 Numeric Results

The total number of data channels (C) is set to be 4 and the queue size (B) is set to be 4. We compare three channel allocation schemes. The first two are BCA and USFCA schemes described in the previous section. The third one is a simple FCFS channel allocation scheme with a FIFO queue of the same size (B). The simple FCFS scheme can also be modeled as a M/M/C/B Markov process, let W_FCFS and T_FCFS denote the mean waiting time and

system time of packets.

The mean service time of one packet ( 1

µ

_P and 1

µ

_nP ) is assumed to be 0.0625 seconds with one timeslot allocated. This represents approximately an average 105 bytes per packet under the GPRS CS-2 coding scheme and is near the average uplink packet sizes. For packet arrival, λ nP is fixed at 32 packets/second and λP varies in the range of 8-32

11

Fig. 2.3: The mean waiting time and system time of uplink packets for λ nP = 32

packets/second and λP= 8-32 packets/second

In Fig. 2.3, the results indicate both BCA and USFCA schemes provide shorter mean waiting time and system time for priority packets than the simple FCFS scheme at the cost of longer mean waiting time and system time for non-priority packets. The improvement and the cost become more significant as the priority traffic increases. In addition, the improvement and the cost of USFCA scheme are more significant than those of BCA scheme. This is because when there is no free channel, USFCA scheme can suspend the transmission of a non-priority packet and start transmitting a new priority packet, but BCA scheme cannot. The improvement on mean waiting time and system time for priority packets over the simple FCFS scheme can be as large as 0.025 seconds when the priority packet arrival rate is the 32 packets/sec, the transmission delay can be greatly reduced to an extend of nearly 72%. This 0.025 seconds difference could be critical for real-time voice communications.

Priority packet arrival rate (packets/sec)

0 0.02 0.04 0.06 0.08 0.1 0.12 8 16 24 32 W a itin g time a n d S ys te m time Tnp_USF Tnp_bm T_FCFS Tp_bm Tp_USF Wnp_USF Wnp_bm W_FCFS Wp_bm Wp_USF

Priority packet arrival rate (packets/sec)

0 0.02 0.04 0.06 0.08 0.1 0.12 8 16 24 32 W a itin g time a n d S ys te m time Tnp_USF Tnp_bm T_FCFS Tp_bm Tp_USF Wnp_USF Wnp_bm W_FCFS Wp_bm Wp_USF Tnp_USF Tnp_bm T_FCFS Tp_bm Tp_USF Wnp_USF Wnp_bm W_FCFS Wp_bm Wp_USF

12

2.5 Conclusions

This chapter, we studied BCA and USFCA schemes that implement priority queuesin the GPRS network. Both schemes provide shorter mean waiting time and system time for priority packets than the simple FCFS scheme at the cost of longer mean waiting time and system time for non-priority packets. In addition, the transmission delay of priority packets using USFCA can be better guaranteed than that of BCA especially when the GPRS traffic is heavy.

13

CHAPTER 3

Uplink Connection Scheduling for Flat-Rate Data

Services in the UMTS Network

3.1 Introduction

The Universal Mobile Telecommunication System (UMTS) using Wideband Code Division Multiple Access (WCDMA) radio technology represents an evolution in terms of capacity, data rates and service capabilities, from the GSM/General Packet Radio Service (GPRS) network [6]. It is an integrated solution for mobile voice and data with wide area coverage and high data rates. The UMTS network can provide packet data rates up to 384 kbps in high mobility situations, and as high as 2 Mbps for stationary users. The packet data usage of current UMTS users is not popular because of the lack of popular mobile data applications and the high cost of data transmission. To attract more packet data users, UMTS operators have begun to provide flat-rate packet services. Flat-rate users pay fixed monthly charge for un-limited data packet transmission. Since usage incurs no extra charge, flat-rate users tend to keep data connections alive longer, and occupy most of the network resources. Without special treatments for different classes of user connections, normal users who are charged by usage may be blocked from accessing the UMTS network.

Since blocked normal user connections result in revenue loss of the UMTS operator, to increase the revenue, normal users should be given priority on transmission. On the other hand, if flat-rate users experience blocked connections frequently, the discontent flat-rate users may switch to other service providers. The loss of flat-rate users leads to revenue loss

14

too. Therefore, a balance needs to be found in allocating radio resources to normal users and flat-rate users. In this chapter, we propose a cost function representing the revenue loss due to both blocked normal users and lost flat-rate users. Since the revenue loss on both situations depends on the blocking probabilities, we investigate scheduling techniques, including queueing, guard channels, preemption and rate-adaptation, to keep the blocking probabilities of normal users and flat-rate users at different levels, and to minimize the cost function, i.e., the revenue loss. We consider the aforementioned four scheduling techniques, because they have been repeatedly used in giving transmission priority in mobile networks. However, no one has investigated the effectiveness of these four scheduling techniques in maximizing the revenue of UMTS operators serving flat-rate and normal users.

Much research has been done on the mobile network in giving transmission priority to a certain type of service. In mobile networks, terminating a handoff call is considered a higher cost than blocking a new call. When a handoff call arrives, but there is no free channel in the cell, the handoff call can be placed in a queue and handoff is delayed until free channels become available [7]. To further give a priority to handoff calls, a small number of free channels called guard channels can be reserved for handoff calls. Guard channels significantly reduce the forced termination probability of handoff calls at the cost of blocking more new calls and reducing the system throughput [8]. To increase the total carried traffic and improve the perceived service quality, Guerin put originating calls in a queue when the network has very few free resources [9]. Zeng, et al, also proposed that both the new and handoff calls can be queued, and showed that the forced termination probability of handoff calls decreased drastically with only a small increase in the blocking probability of new calls [10]. For integrated voice and data communications, Zeng, et al, presented a system with two queues for handoff calls, one for voice and the other for data. Their results showed that the forced termination probability of voice handoff calls and the average transmission delay of data connections decreased by increasing the size of handoff queues [11]. Leong, et al, presented a

15

system with two buffers for data calls, one for new data call and the other for handoff. Their results indicated that the Quality of Service (QoS) can be guaranteed for both voice and data services in a multi-cell environment [12].

Preempting a low priority call to free radio resources for high priority calls is another effective way to ensure transmission priority. However, this approach usually preempts data calls only, because cutting off voice communications can be very annoying to the users. High priority of real-time traffic, such as voice and video, can preempt non-real-time traffic (data). Several researchers have shown that the preemption of non-real-time data can guarantee QoS for real-time classes, and achieve high channel utilization [13-14]. Kim, et al, proposed that high priority voice calls can preempt low priority voice calls. Voice calls that have low SIR and long duration are considered low priority calls, which can be preempted to improve the entire network performance [15].

Sub-rating current calls to free radio resources for new or handoff calls is another way to reduce blocking probabilities. A serving full-rate channel can be temporarily divided into two half-rate channels when the network is fully utilized; one to serve the existing call and the other to serve the handoff call [16]. Chen, et al, studies GPRS networks where a data session can occupy more than one GPRS data channel. When there are no free channels upon the arrival of a voice call, one slot of an existing multi-slot GPRS data session is de-allocated for the new voice arrival [17]. Their results show the voice blocking probability can be greatly reduced, especially at high GPRS traffic load.

Most of the researches focus on reducing the blocking and forced termination probabilities of high-priority connections. However, very few studies have been done on maximizing the operator revenue for mobile networks serving flat-rate users, as well as normal users. In this chapter, we investigate combinations of the scheduling techniques aforementioned to maximize the operator revenue. We propose a cost function that represents the revenue loss of service providers providing both flat-rate and per-packet charging services.

16

An iterative algorithm has been developed to determine the optimal number of guard channels and the best combination of scheduling techniques in minimizing the revenue loss. Our study may be valuable for UMTS operators in serving flat-rate users. The notations we use in this chapter are listed in Table 3.1.

Table 3.1 The usage of the system notations in uplink connection scheduling for Flat-Rate Users

Notation Meaning

B The size of the NUC waiting queue

C The cost function

Cf The monthly revenue loss due to lost flat rate users

Cmin The minimum value of the cost function

Cn The monthly revenue loss due to blocked NUCs

G The number of guard channels

Gopt The optimum number of guard channels

LQN The average NUC queue lengths

LQF The average FRUC queue length

NF The number of full-rate connections

) (

*

y

NF The maximum number of full-rate connections

17

NFF(i,j,k) The number of full-rate FRUCs in state (i, j, k)

NH The number of half-rate connections

) z ( N*

H The maximum number of half-rate connections

when there are zfull-rate connections

NHF(i,j,k) The number of half-rate FRUCs in state (i, j, k) NHF+FF(i,j,k) The total number of full-rate and half-rate FRUCs

in state (i, j, k)

PBF The blocking probability of FRUCs

PBN The blocking probabilities of NUCs

PF The probability that the first events occurs to a serving half-rate FRUC is being full-rated PFC The probability that the first events occurs to a

serving full-rate FRUC is completion

PFP The probability that the first events occurs to a full-rate serving FRUC is being preempted PFPrm The probability that a serving full-rate FRUC is

preempted before its completion

PFS The probability that a serving full-rate FRUC is sub-rated before its completion or preemption PFST PFST =1−PFC−PS −PFP

Pi,j,k The stationary state probability of the network in state (i,j,k)

PS The probability that the first events occurs to a serving full-rate FRUC be being sub-rated PSC The probability that the first events occurs to a

18

serving half-rate FRUC is completion

PSP The probability that the first events occurs to a serving half-rate FRUC is being preempted PSPrm The probability that a serving half-rate FRUC is

preempted before its completion PSST PSST =1−PSC−PF−PSP

PQF The queueing probability of FRUCs

PQN The queueing probability of NUCs

SAll Scheduler with guard channels, waiting queues, rate adaptation, and preemption scheduler SG the set of all existing transition states of SAll SNPrm Scheduler without preemption

SNRA Scheduler without rate adaptation

SNWQ Scheduler without the NUC waiting queues

TX The average transmission rate of serving FRUCs

Q The size of the FRUC waiting queue

WTF The waiting time of FRUCs

WTN The waiting time of NUCs

α

The cost weighting factor of flat-rate users F

α

The activity factor of full-rate connections H

19

β The departure threshold of FRUC blocking

probability

F

δ

The nominal capacity of a full-rate connection H

δ

The nominal capacity of a half-rate connection f

λ

The arrival rate of FRUCs

n

λ The arrival rate of NUCs

f

/

µ

1 The average service time of full-rate FRUCs

n

/

µ

1 The average service time of NUCs

k , j , i

θ

An indicator to indicate whether state (i,j,k) exists or not

n

ρ

The traffic load of NUCs

ζ

The inter-cell interference factor for a cell

) N , N ( F H

Ω

The total transmission power received by the RNC in a cell

3.2 System Models and Assumptions

A UMTS network consists of three interacting domains: Core Network (CN), UMTS Terrestrial Radio Access Network (UTRAN) and mobile stations (MS). The UTRAN provides the air interface access method for MSs [18]. A Base Station is referred to as Node B; the control node for a group of Node Bs is called a Radio Network Controller (RNC). Wideband CDMA technology was selected to be the air interface of the UTRAN. To be specific, we study the Frequency Division Duplex (FDD) WCDMA operation in this chapter. A RNC can allocate a physical dedicated radio channel (D_CH) to an MS by through a RAB assignment procedure [18-20]. Fig. 3.1 depicts the message flow of a D_CH assignment procedure. In Step 1, an MS establishes a Radio Resource Control (RRC) connection with the

20

RNC before creating a Packet Data Protocol (PDP) context between the MS and the GGSN. In Step 2, the MS sends an “Activate PDP Context Request” message to the SGSN with a QoS element indicating service class (conversional, streaming, interactive or background data). In Step 4, the SGSN sends a “RAB assignment request” message with RAB parameters, which will be described in more details later, to the RNC to establish a RAB connection between the MS and SGSN. After the D_CH is established in Step 5, the MS can start to transmit/receive packets to/from the CN in Step 6. When necessary, the RNC can instruct the MS that packet transmission of the connection should be stopped, continued or change the transmission rate on its assigned D_CH by a Radio Bearer (RB) reconfiguration procedure as indicated in Step 7. The MS should comply with the instructions. After the MS completes transmission, the RB and RRC of the MS can be released in Steps 8 and 9.

Fig. 3.1: The Radio Access Bearer (RAB) assignment procedure

“Connection”

MS RNC SGSN GGSN

1.RRC request

2.Activate PDP Context request (QoS)

3.Create PDP Context (QoS)

4. RAB assignment request (RAB parameters)

5.Radio Bearer setup (D_CH)

6.Transmit data between UE and PCN

8.Radio Bearer release (D_CH)

9.RRC release

7.Radio Bearer reconfiguration (stop/continue/rate-adapted) Node B IuPS Gn Uu Iub “Connection” MS RNC SGSN GGSN 1.RRC request

2.Activate PDP Context request (QoS)

3.Create PDP Context (QoS)

4. RAB assignment request (RAB parameters)

5.Radio Bearer setup (D_CH)

6.Transmit data between UE and PCN

8.Radio Bearer release (D_CH)

9.RRC release

7.Radio Bearer reconfiguration (stop/continue/rate-adapted)

Node B IuPS Gn

21

The RAB parameters sent from the SGSN to the RNC in Step 4 can be used to instruct the RNC the scheduling policy of the data connection. The parameters include a Priority Level element, a Pre-emption Capability element indicating the capability to preempt lower priority RABs, a Pre-emption Vulnerability element indicating whether the DCH is vulnerable to be preempted or not, and a Queuing Allowed element indicating whether the RAB request can be queued. In addition, Maximum Bit Rate and Guaranteed Bit Rate elements indicate the transmission rate of MSs. These RAB parameters can be used to instruct the RNC how to schedule the packet transmission

A user connection starts at the establishment of a RRC between the MS and the RNC, as shown in Step 1, Fig. 3.2, and ends at the disconnection of the RNC. We assume there are two types of user connections in the UMTS network; Normal User Connections (NUCs), which are assigned a higher priority in transmission, and Flat-Rate User Connections (FRUCs), which may be sub-rated or suspended when the network traffic load is high. When a connection is sub-rated, its transmission rate and transmission power can be reduced, and thus transmission power allowance is released for other connections. Since NUCs are charged by the volume of packet transmission, a NUC tends to be shorter, such as sending an e-mail or uploading short files. On the other hand, FRUCs pay fixed monthly fee no matter how many packets they transmit, a FRUC is generally longer, such as playing on-line games and using peer-to-peer applications.

In UMTS R99 network, the uplink data transmission can only be scheduled on connection level, but not on packet level. This is because after DCHs are allocated to MSs, the MSs can start or pause data transmission anytime without notifying the RNC. However, the RNC can suspend or sub-rate the uplink connection as we have described. On the other hand, downlink data transmission can be scheduled on packet level, because all downlink packets are stored and forwarded by the RNC. The RNC can determine priorities in forwarding different classes of packets. As a result, uplink and downlink transmissions may require

22

different scheduling techniques. In this chapter, we consider the scheduling for uplink data connections only. We use the CDMA uplink soft-capacity model to estimate the uplink total bandwidth of a cell in system.

3.2.1

CDMA Uplink Capacity Model

The capacity of a CDMA network is not fixed; it has so-called “soft capacity”. Since a FRUC can be sub-rated, we consider two transmission rates of data services from MSs, full-rate and half-rate data connections. We can obtain the limit on the total transmission power received by the Node B in a cell from Equation (1) [21].

α

F and

α

H denote the

activity factor of full-rate and half-rate data connections in a cell, respectively. N and F H

N denote the numbers of MSs using full-rate and half-rate data connections,

respectively. δF denotes the nominal capacity of a full-rate data connection, i.e., the

portion of total transmission power received by the Node B in a cell; δH denotes that of a

half-rate data connection [22].

ς

is the inter-cell interference factor for a cell which can be obtained from measurements [23].

( ) N N ) N , N ( F H F F F H H H ζ δ α δ α Ω + < × × + × × = 1 1 (3.1)

From Equation (3.1), we can obtain the Pole capacity of MSs using full-rate and half-rate data connections in a cell as in Equation (3.2). N*_F

( )

y denotes the maximum number of full-rate serving MSs in a cell when there are y half-rate serving MSs and

( )

z

N*_H denotes the maximum number of half-rate serving MSs in a cell when there are z full-rate serving MSs. In particular, N*F

( )

0 denotes the maximum number of full-rate serving MSs in a cell, and N*H

( )

0 half -rate serving MSs.

23 1 1 1 1       × × + × × ≥ + =       × × + × × ≥ + = ) N z ( ) ( N max ) z ( N ) y N ( ) ( N max ) y ( N H H H F F H * H H F F F F * F H _ζ α δ α δ δ α δ α ζ (3.2)

In the analysis below, we assume the spread spectrum bandwidth (W) of the WCDMA network is 5 MHz, the full-rate data transmission is 128kbps and the half-rate is 64kbps. The two data rates are the default uplink data rates provided in CHT UMTS R99 network. According to the 3GPP specification [24], full-rate and half-rate data transmissions have different Signal-Interference-Ratio (SIR) requirements to achieve Block Error Rate (BLER) <10−2 in multipath fading conditions; for full-rate it is 8.4

dB and half-rate 9.2 dB. From the desired SIR, we can obtain the nominal capacity δF

= 0.177and δH =0.106. The activity factor for data services (

α

F and

α

H) is assumed

to be 0.5 in busy hour, and the inter-cell interference factor (

ς

) is assumed to be 0.1. These assumptions follow those in [23].

From Equation (2), we can obtain *

( )

0

F

N = 10 and N*H( )0 = 17. Note that

( )

0

* H

N is less than twice of *

( )

0

F

N because more number of MSs transmitting leads to more signal interference. In other words, half-rate transmission is less efficient in using radio bandwidth.

3.2.2

A Scheduler with all four features

The queueing model of the connection scheduler that implements waiting queues (WQ), guard channels (GC), preemption and rate-adaptation on the RNC is depicted in Fig. 3.2. There are two waiting queues; one for new NUCs, and the other for new and preempted FRUCs. When an on-going up-link connection is put in a waiting queue, the Node B instructs the MS to stop packet transmission. Since there is no packet transmission, no storage space on Node-B is needed for the up-link packets of a queued