radio network gprs qos

TALLINNA TEHNIKAÜLIKOOL

Raadio- ja sidetehnika instituut

Kood: IRT84 LT

GPRS TEENUSTE KVALITEETI MÕJUTAVAD RAADIOVÕRGU SEADED

The impact of radio network factors on the QoS of GPRS applications

Aleksandr Brizmer

Töö on tehtud telekommunikatsiooni õppetooli juures Juhendaja Avo Ots Kaitsmine toimub raadio- ja sidetehnika instituudi kaitsmiskomisjonis Autor taotleb tehnikateaduste magistri nimetust Esitatud: 25.01.2005 Kaitsmine: 09.02.2005

Tallinn 2005


SUMMARY The purpose of this work is to describe the impact of radio network factors (cell configuration, traffic load and interference) on end-user quality of service of GPRS applications (WAP and HTTP) in a mobile network. End-user quality of service is expressed through a set of parameters defined by ITU-T E.800 Recommendation. These parameters were measured during the tests that were performed in one cell of a GSM network. The results of the tests with their analysis are presented. The work is divided into two main parts. The first part contains the theoretical background of GPRS service, QoS aspects and GPRS QoS-affecting factors in a mobile network. The second part contains the information about tests: measurement system overview, tests scenarios and results with their analysis. This work is comprised of 85 pages, including 40 figures and 12 tables. Essential keywords pertaining this work are:

• GPRS • QoS • QoS parameters • QoS aspects • Network performance • PDCH configuration

Alexander Brizmer 2


DIE KURZFASSUNG Der Zweck von dieses Arbeit ist die Aufswirkung der GSM-Radionetzwerkfaktoren (die Funkzellekonfiguration, die Verkehrsauslastung in der Funkzell, die Störungsgeräusche) auf die Endbenutzerdienstgüte die GPRS-Anwendungen (WAP und HTTP) zu beschreiben. Die Endbenutzerdienstgüte ist durch die Anlage der Parametren ausgedrückt. Diese Parameter sind bei ITU-T E.800 Empfellung bestimmt. Die Parameter wurden in eine Funkzelle der GSM Netzwerk abgemessen, und die Ergebnisse mit der weiterer Analyse wurden aufgezeigt. Die Arbeit ist in zwei Hauptaneile abgeteilt. Das erstes Anteil beinhaltet den theoretischen Hintergrund der GPRS Dienstes, die Dienstgüteaspekte und die GSM-Radionetzwerkfaktoren die der Aufswirkung auf der GPRS Dienstgüte haben. Das zweites Anteil beinhaltet der Information über die Testen: die Übersicht des Abmessungsystemes, die Testszenarios und die Ergebnisse mit der Analyse. Die Arbeit besteht aus 85 Blätte und beinhaltet 40 Bilder und 12 Tabellen. Die wichtigste Schlüsselwörter, das diese Arbeit betreffen, sind:

• GPRS • GPRS Dienstgüte (GPRS QoS) • QoS Parameter • QoS Aspekte • Die Netzwerkleistung • PDCH Konfiguration

Alexander Brizmer 3


REFERAAT Käesoleva töö eesmärk on kirjeldada raadiovõrgu seadete (kärje konfiguratsioon, kärje koormus ja interferents) mõju GPRS mobiilsidevõrgu rakenduste (WAP ja HTTP) teenusekvaliteedile. Lõppkasutaja teenuse kvaliteet on väljendatud ITU-T E.800 soovitustega sätestatud parameetrite hulga kaudu. Neid parameetreid mõõdeti GSM võrgu kärje teeninduspiirkonnas sooritatud testide käigus. Töös esitatakse testide tulemused ja nende analüüs. Töö on jagatud kaheks osaks. Esimene osa sisaldab GPRS’i ülevaadet, teenusekvaliteedi aspekte ja GPRS’i teenusekvaliteeti mõjutavate raadiovõrgu seadete tutvustust. Teine osa sisaldab informatsiooni testide kohta: mõõtmissüsteemi ülevaade, testide stsenaariumid ning tulemused koos nende analüüsiga. Töö maht on 85 lehekülge, sisaldades 40 joonist ja 12 tabelit. Käesoleva töö võtmesõnad on järgmised:

• GPRS • Teenusekvaliteet (QoS) • QoS parameetrid • QoS aspektid • Võrgu jõudlus • PDCH konfiguratsioon

Alexander Brizmer 4


PREFACE This work covers the impact of radio network factors on the Quality of Service (QoS) of GPRS applications in mobile networks. The reason why I have chosen this topic is in the growing demand on GPRS service and GPRS-based Internet applications. This demand sets certain requirements for the QoS of these applications. These requirements are expressed by sets of QoS parameters. ITU-T Recommendation E.800 and ETSI Technical Standard 102 250-2 V1.2.1 define these parameters for GPRS. There are several factors that affect the QoS of GPRS applications in mobile networks. Some of them are related to radio network (e.g. co-channel interference, traffic load and packet data channel configuration in a cell). The goal of this work is to define the impact of radio network factors on the QoS of GPRS applications in mobile networks. For this purpose, a number of tests have been conducted in the cellular network of Estonian GSM operator AS EMT. Two most commonly used applications- WAP and Web browsing (HTTP) were tested in different cell configuration and traffic load conditions with regards to the level of co-channel interference. The tests were performed using a complex measurement system, which consisted of end-user equipment and special software tools. The results of the tests were analysed in order to create a vision of the impact of every radio network factor on the end-user QoS of these two applications. This work has practical meaning, showing the impact of radio network factors on the performance and quality of packet data communication in a real cellular network. It may be useful for mobile operators who want to improve the performance of their GPRS services. I would like to thank my colleagues at EMT- Mr. Ermo Polma, Mr. Andres Laidvee and Mr. Allan Soikonen, who helped me to set up the tests in EMT network and consulted me in tricky theoretical aspects of this work. I would also like to thank my supervisor at Tallinn Technical University, Mr. Avo Ots, who helped me to make a right choice of a topic for Master of Science Degree work and whose help in theoretical part of the work was very essential. Alexander Brizmer, 13 January 2005

Alexander Brizmer 5


TABLE OF CONTENTS

SUMMARY ........................................................................................................................ 2

DIE KURZFASSUNG ...................................................................................................... 3

REFERAAT....................................................................................................................... 4

PREFACE ......................................................................................................................... 5

INDEX OF FIGURES..................................................................................................... 10

INDEX OF TABLES....................................................................................................... 12

ACRONYMS ................................................................................................................... 13

DEFINITIONS ................................................................................................................. 15

INTRODUCTION ............................................................................................................ 17

1. GPRS OVERVIEW................................................................................................. 19

1.1 INTRODUCTION TO GPRS ...................................................................................... 19 1.2 GPRS ARCHITECTURE AND INTERFACES .............................................................. 20

1.2.1 GSM network nodes .................................................................................... 21 1.2.1.1 Mobile Station ........................................................................................ 21 1.2.1.2 Base Transceiver Station .................................................................... 22 1.2.1.3 Base Station Controller ........................................................................ 22 1.2.1.4 Mobile Switching Center ...................................................................... 22 1.2.1.5 Visitor Location Register ...................................................................... 23 1.2.1.6 Home Location Register ...................................................................... 23

1.2.2 GPRS nodes ................................................................................................. 23 1.2.2.1 Serving GPRS support node............................................................... 23 1.2.2.2 Gateway GPRS support node ............................................................ 23

1.2.3 GPRS interfaces........................................................................................... 24 1.2.4 GPRS protocols ............................................................................................ 25

1.3 GPRS MOBILITY MANAGEMENT ............................................................................. 25 1.3.1 MS Attach and detach ................................................................................. 26 1.3.2 PDP Context activation/deactivation ......................................................... 27

1.3.2.1 PDP Context activation ........................................................................ 27 1.3.2.2 PDP Context Deactivation ................................................................... 28

1.3.3 Data transfer ................................................................................................. 28 1.3.4 Cell Reselection............................................................................................ 28

1.4 GPRS APPLICATIONS ............................................................................................ 29

2. QUALITY OF SERVICE GENERAL CONCEPTS............................................ 30

2.1 DEFINITIONS OF QOS............................................................................................. 30 2.2 QUALITY CYCLE. OFFERED VERSUS DELIVERED QOS ......................................... 30 2.3 QOS LEVELS........................................................................................................... 31 2.4 QOS ASPECTS AND PARAMETERS ......................................................................... 32

2.4.1 GPRS QoS aspects ..................................................................................... 33

Alexander Brizmer 6


2.4.1.1 Service Accessibility ............................................................................. 33 2.4.1.2 Service Retainability ............................................................................. 33 2.4.1.3 Service Integrity .................................................................................... 33

2.4.2 GPRS QoS parameters............................................................................... 34 2.4.2.1 GPRS Accessibility Failure Ratio ....................................................... 35 2.4.2.2 GPRS End-to-end Access Time ......................................................... 37 2.4.2.3 GPRS Mean User Data Rate .............................................................. 37 2.4.2.4 GPRS Data Transfer Cut-off Ratio..................................................... 38 2.4.2.5 Service Round Trip Time ..................................................................... 38

2.5 QOS REQUIREMENTS FOR GPRS-BASED APPLICATIONS ..................................... 38 2.5.1 Reliability ....................................................................................................... 38 2.5.2 Delay (latency and jitter) ............................................................................. 39

3. RADIO NETWORK FACTORS THAT AFFECT GPRS QOS ........................ 40

3.1 BANDWIDTH AVAILABILITY ...................................................................................... 41 3.1.1 GPRS channels ............................................................................................ 41

3.1.1.1 GPRS physical channels ..................................................................... 41 3.1.1.2 GPRS logical channels ........................................................................ 41

3.1.2 PDCH configuration ..................................................................................... 42 3.1.2.1 Dynamic PDCH configuration ............................................................. 42 3.1.2.2 Static PDCH configuration................................................................... 42 3.1.2.3 Combined PDCH configuration .......................................................... 42

3.2 INTERFERENCE ....................................................................................................... 43 3.2.1 Channel coding............................................................................................. 43 3.2.2 Link adaptation ............................................................................................. 44 3.2.3 Coding schemes for retransmissions. Selective ARQ. .......................... 44

4. TESTED SYSTEM AND TEST SETUP.............................................................. 45

4.1 A SHORT OVERVIEW OF TESTED NETWORK ........................................................... 45 4.1.1 Overview of EMT GPRS structure............................................................. 45

4.2. MEASUREMENT SYSTEM ....................................................................................... 46 4.2.1 TEMS™ Investigation GSM – a short description .................................. 47 4.2.2 Additional tools: SGSN logging and acceSS7 monitoring. .................... 48

4.3 TEST SCENARIOS.................................................................................................... 49 4.3.1 Tested cell: PDCH configuration, cell capacity and load ....................... 49 4.3.2 Tested services ............................................................................................ 50 4.3.3 Tested parameters ....................................................................................... 50

5. MEASUREMENTS OF QOS PARAMETERS................................................... 52

5.1 KPI MEASUREMENTS: HTTP, NORMAL CELL TRAFFIC LOAD, DYNAMIC PDCH CONFIGURATION ............................................................................................................ 52

5.1.1 GPRS Accessibility Failure Ratio .............................................................. 52 5.1.2 GPRS End-to-end Access Time ................................................................ 52 5.1.3 GPRS Mean User Data Rate ..................................................................... 53 5.1.4 GPRS Data Transfer Cut-off Ratio ............................................................ 53 5.1.5 Round trip time and packet loss ................................................................ 53

Alexander Brizmer 7


5.2 KPI MEASUREMENTS: HTTP, HIGH CELL TRAFFIC LOAD, DYNAMIC PDCH CONFIGURATION ............................................................................................................ 53

5.2.1 GPRS Accessibility Failure Ratio .............................................................. 53 5.2.2 GPRS End-to-end Access Time ................................................................ 54 5.2.3 GPRS Mean User Data Rate ..................................................................... 54 5.2.4 GPRS Data Transfer Cut-off Ratio ............................................................ 54 5.2.5 Round trip time and packet loss ................................................................ 54

5.3 KPI MEASUREMENTS: HTTP, HIGH CELL TRAFFIC LOAD, COMBINED PDCH CONFIGURATION ............................................................................................................ 54

5.3.1 GPRS Accessibility Failure Ratio .............................................................. 55 5.3.2 GPRS End-to-end Access Time ................................................................ 55 5.3.3 GPRS Mean User Data Rate ..................................................................... 55 5.3.4 GPRS Data Transfer Cut-off Ratio ............................................................ 55 5.3.5 Round Trip Time and Packet Loss ............................................................ 55

5.4 KPI MEASUREMENTS: WAP, NORMAL CELL TRAFFIC LOAD, DYNAMIC PDCH CONFIGURATION ............................................................................................................ 56

5.4.1 GPRS Accessibility Failure Ratio .............................................................. 56 5.4.2 GPRS End-to-end Access Time ................................................................ 56 5.4.3 GPRS Mean User Data Rate ..................................................................... 56 5.4.4 GPRS Data Transfer Cut-off Ratio ............................................................ 56 5.4.5 Round Trip Time and Packet Loss ............................................................ 56

5.5 KPI MEASUREMENTS: WAP, HIGH CELL TRAFFIC LOAD, DYNAMIC PDCH CONFIGURATION ............................................................................................................ 57

5.5.1 GPRS Accessibility Failure Ratio .............................................................. 57 5.5.2 GPRS End-to-end Access Time ................................................................ 57 5.5.3 GPRS Mean User Data Rate ..................................................................... 57 5.5.4 GPRS Data Transfer Cut-off Ratio ............................................................ 57 5.5.5. Round Trip Time and Packet Loss ........................................................... 58

5.6 KPI MEASUREMENTS: WAP, HIGH CELL TRAFFIC LOAD, COMBINED PDCH CONFIGURATION ............................................................................................................ 58

5.6.1 GPRS Accessibility Failure Ratio .............................................................. 58 5.6.2 GPRS End-to-end Access Time ................................................................ 58 5.6.3 Mean User Data Rate .................................................................................. 58 5.6.4 Data Transfer Cut-Off Ratio........................................................................ 59 5.6.5 Round Trip Time and Packet Loss ............................................................ 59

5.7 COMPARISON OF TESTS’ RESULTS......................................................................... 60 5.7.1 HTTP .............................................................................................................. 60

5.7.1.1 Accessibility Failure Ratio.................................................................... 60 5.7.1.2 End-to-end Access Time ..................................................................... 61 5.7.1.3 Mean User Data Rate .......................................................................... 62 5.7.1.4 Data Transfer Cut-off Ratio ................................................................. 62 5.7.1.5 Round Trip Time and Packet Loss..................................................... 63

5.7.2 WAP ............................................................................................................... 64 5.7.2.1 Accessibility Failure Ratio.................................................................... 64 5.7.2.2 End-to-end Access Time ..................................................................... 65 5.7.2.3 Mean User Data Rate .......................................................................... 66

Alexander Brizmer 8


5.7.2.4 Data Transfer Cut-off Ratio ................................................................. 66 5.7.2.5 Round Trip Time and Packet Loss Ratio .......................................... 67

CONCLUSIONS ............................................................................................................. 68

APPENDIX A: TWO METHODS OF MEAN DATA RATE MEASUREMENTS 72

APPENDIX B: BLOCK STRUCTURES AND PARAMETERS OF GPRS CHANNEL CODING SCHEMES ................................................................................. 74

APPENDIX C: LINK ADAPTATION ALGORITHMS .......................................... 76

C.1 ESTIMATED C/I ALGORITHM: ................................................................................. 76 C.2 BLOCK ERROR RATE ALGORITHM: ......................................................................... 76

APPENDIX D: RETRANSMISSION WITH SELECTIVE ARQ.......................... 78

APPENDIX E: OVERVIEW OF EDGE (EGPRS) ................................................ 80

E.1 TECHNICAL DIFFERENCES BETWEEN GPRS AND EGPRS .................................. 81 E.1.1 EDGE modulation technique...................................................................... 82 E.1.2 Coding schemes .......................................................................................... 82 E.1.3 Packet handling............................................................................................ 83 E.1.4 Addressing window...................................................................................... 84 E.1.5 Interleaving ................................................................................................... 84 E.1.6 EGPRS link controlling function ................................................................ 84

E.2 IMPACT OF EGPRS ON EXISTING GSM/GPRS NETWORKS................................ 85

Alexander Brizmer 9


INDEX OF FIGURES Figure 1.1 The evolution of Mobile Data Services (by the year 2002) [2] ............. 20 Figure 1.2 GSM network with GPRS functionality .................................................... 20 Figure 1.3 GPRS architecture and interfaces ............................................................ 21 Figure 1.4 GPRS protocol stack [1]............................................................................. 25 Figure 1.5 GPRS MS states ......................................................................................... 26 Figure 1.6 MS Attach ..................................................................................................... 26 Figure 1.7 PDP Context Activation procedure........................................................... 28 Figure 1.8 MS-initiated PDP Context deactivation procedure ................................ 28 Figure 2.1 The quality cycle [10] .................................................................................. 31 Figure 2.2 A three-layer model of QoS aspects and parameters [12] ................... 32 Figure 2.3 KPIs, PIs, and timeouts for all supported data services except WAP

and MMS [13].......................................................................................................... 34 Figure 2.4 KPIs, PIs, and timeouts for WAP over GPRS [13]................................. 35 Figure 3.1GSM physical channels [14] ....................................................................... 41 Figure 3.2 Allocation of packet data channels in GSM network [17] ..................... 41 Figure 3.3 Combined static and dynamic configuration of PDCH channels in a

cell [17]..................................................................................................................... 43 Figure 3.4 Throughput versus carrier to interference ratio [21] .............................. 44 Figure 4.1 A simplified view of GPRS structure of the tested network .................. 46 Figure 4.2 Measurement system for GPRS active tests.......................................... 47 Figure 4.3 TEMS™ Investigation GSM user interface ............................................. 48 Figure 5.1 Accessibility Failure Ratio for HTTP (charts): general values (A) and

performance indicators’ values (B) ...................................................................... 61 Figure 5.2 GPRS End-to-end Access Time for HTTP (charts): general values (A)

and performance indicators’ values (B) .............................................................. 62 Figure 5.3 Mean User Data Rate for HTTP (chart)................................................... 62 Figure 5.4 Data Transfer Cut-off Ratio for HTTP (chart) ......................................... 63 Figure 5.5 RTT (A) and Packet Loss Ratio (B) for HTTP (charts).......................... 63 Figure 5.6 Accessibility Failure Ratio for WAP (charts): general values (A) and

performance indicators’ values (B) ...................................................................... 65 Figure 5.7 GPRS End-to-end Access Time for WAP (charts): general values (A)

and performance indicators’ values (B) .............................................................. 65 Figure 5.8 Mean User Data Rate for WAP (chart) .................................................... 66 Figure 5.9 Data Transfer Cut-off Ratio for WAP (chart)........................................... 66 Figure 5.10 RTT (A) and Packet Loss Ratio (B) for WAP (charts)......................... 67 Figure A.1 Measurement of Mean Data Rate for HTTP, Method A [12] ............... 72 Figure A.2 Measurement of Mean Data Rate for HTTP, Method B [12] ............... 73 Figure B.1 Data flow segmentation between layers [19] ......................................... 74 Figure B.2 Encoding of GPRS data packets [14] ...................................................... 74 Figure B.3 Radio block structures for GPRS coding schemes: a) CS1-CS3 b)

CS4 [19] ................................................................................................................... 75 Figure C.1 an example of BLER-based algorithm [19] ............................................ 77 Figure D.1 Retransmission using Selective Repeat [28] ......................................... 78

Alexander Brizmer 10


Figure D.2 An example of retransmission with selective ARQ [28]........................ 79 Figure E.1 EGPRS introduces changes to GPRS only on the base station system

part of the network [29].......................................................................................... 81 Figure E.2 I/Q diagram showing EDGE modulation benefits [29] .......................... 82 Figure E.3 Coding schemes for GPRS and EGPRS (user data rate per time slot)

[29] ............................................................................................................................ 83



INDEX OF TABLES Table 1.1 GPRS interfaces ........................................................................................... 24 Table 3.1 GPRS logical channels [14] ........................................................................ 42 Table 5.1 Measured RTT and Packet loss, HTTP, dynamic PDCH configuration,

normal cell traffic load............................................................................................ 53 Table 5.2 Measured RTT and Packet loss, HTTP, dynamic PDCH configuration,

high cell traffic load ................................................................................................ 54 Table 5.3 Measured RTT and Packet loss, HTTP, combined PDCH configuration,

high cell traffic load ................................................................................................ 55 Table 5.4 Measured RTT and Packet loss, WAP, dynamic PDCH configuration,

normal cell traffic load............................................................................................ 56 Table 5.5 Measured RTT and Packet loss, WAP, dynamic PDCH configuration,

high cell traffic load ................................................................................................ 58 Table 5.6 Measured RTT and Packet loss, WAP, combined PDCH configuration,

high cell traffic load ................................................................................................ 59 Table 5.7 Comparison of key performance indicators, measured for HTTP ........ 60 Table 5.8 Comparison of key performance indicators, measured for WAP.......... 64 Table B.1 Parameters of GPRS channel coding schemes [30].............................. 75 Table E.1 Comparison of GPRS and EDGE technical data [29] ............................ 81



ACRONYMS 3GPP – Third Generation Partnership Project ANSI – American National Standard Institute APN – Access Point Node ARQ – Automatic Repeat Request BSC – Base Station Controller BSS – Base Station System BSSGP – Base Station System GPRS Protocol BTS – Base Transceiver Station CIR – Carrier-to-Interference Ratio CS – Coding Scheme DMS – Degrees-Minutes-Seconds EDGE – Enhanced Data rates for Global Evolution EGPRS – Enhanced GPRS ERP – Enterprise Resource Planning ETSI – European Telecommunications Standard Institute GGSN – Gateway GPRS Support Node GMSK – Gaussian Minimum Shift Keying GPRS – General Packet Radio Service GRE – General Routing Encapsulation GSM – Global System for Mobile Communication GTP – GPRS Tunnelling Protocol HLR – Home Location Register HSCSD – High Speed Circuit-Switched Data HTTP – Hyper Text Transfer Protocol IP – Internet Protocol IPSec – IP Security protocol ITU – International Telecommunications Union ITU-T – ITU standardization sector KPI – Key Performance Indicator LA – Link Adaptation LAN – Local Area Network LLC – Logical Link Control MAC – Media Access Control MAP – Mobile Application Part MIF – MapInfo data Interchange Format MMS – Multimedia Messaging Service MS – Mobile Station MSC – Mobile Switching Center MT – Mobile Terminal OSS – Operation and Support System PCU – Packet Control Unit PDCH – Packet Data Channel PDN – Packet Data Network PDU – Packet Data Unit PI – Performance Indicator PING – Packet Internet Groper PSK – Phase Shift Keying PTM – Point-to-Multipoint QoS – Quality of Service RADIUS - Remote Authentication Dial In User Service RAI – Routing Area Identifier RF – Radio Frequency



RLC – Radio Link Control RTT – Round Trip Time SGSN – Serving GPRS Support Node SMS – Short Message Service SNDCP – Sub-Network Dependent Convergence Protocol SS7 – Signalling System 7 TBF – Temporary Block Flow TCH – Traffic Channel TCP – Transmission Control Protocol TDMA – Time Division Multiple Access TE – Terminal Equipment UDP – User Datagram Protocol WAP – Wireless Application Protocol WCDMA – Wideband Code Division Multiple Access VLR – Visitor Location Register



DEFINITIONS Application - software that performs a specific task or function (Federal Standard 1037C: Glossary of Telecommunications). Backbone - the high-traffic-density connectivity portion of any communications network. In packet-switched networks, a primary forward-direction path traced sequentially through two or more major relay or switching stations. In packet-switched networks, a backbone consists primarily of switches and interswitch trunks (ATIS Telecom Glossary, 2000). Bandwidth - the difference between the limiting frequencies within which performance of a device, in respect to some characteristic, falls within specified limits (ATIS Telecom Glossary, 2000). Channel - a single path provided by a transmission medium via either (a) physical separation, such as by multipair cable or (b) electrical separation, such as by frequency- or time-division multiplexing (ATIS Telecom Glossary, 2000). Connection - an association established between functional units for conveying information (ATIS Telecom Glossary, 2000). Data - representation of facts, concepts, or instructions in a formalized manner suitable for communication, interpretation, or processing by humans or by automatic means. Any representations such as characters or analogue quantities to which meaning is or might be assigned (ATIS Telecom Glossary, 2000). Data Transmission - the sending of data from one place to another by means of signals over a channel (ATIS Telecom Glossary, 2000). Delay - the amount of time by which an event is retarded (ATIS Telecom Glossary, 2000). Failure - a temporary or permanent termination of the ability of an entity to perform its required function (Federal Standard 1037C: Glossary of Telecommunications). Fault - an accidental condition that causes a functional unit to fail to perform its required function (Federal Standard 1037C: Glossary of Telecommunications). Functional unit - an entity of hardware, software, or both, capable of accomplishing a specified purpose (ATIS Telecom Glossary, 2000). Network - the totality of the hardware and software infrastructure to provide specified services (University of the Witwatersrand, Johannesburg, Glossary of Telecommunications). Network Congestion - a state of overload within a network, where there is a risk of traffic loss or service degradation (National Cable and Telecommunications Association). Network Performance - The qualitative level at which a network fulfills its function (Federal Standard 1037C: Glossary of Telecommunications). Node - in network topology, a terminal of any branch of a network or an interconnection common to two or more branches of a network. In a switched network, one of the switches forming the network backbone (ATIS Telecom Glossary, 2000). Protocol - a set of messages by which two entities communicate together with a specification of the expected behaviour of an entity as a result of receiving a message. Also used in Internet



standards for a specification not involving the transfer of messages between remote computers (University of the Witwatersrand, Johannesburg, Glossary of Telecommunications). Quality of Service - the performance specification of a communications channel or system (Federal Standard 1037C: Glossary of Telecommunications). Reliability - the probability that a functional unit will perform its required function for a specified interval under stated conditions (ATIS Telecom Glossary, 2000). Service - in general the offerings to users of a telecommunications or information service provider that are manageable, billable and have stated performance. In the context of a protocol, a service is offered by a lower layer to a higher layer (University of the Witwatersrand, Johannesburg, Glossary of Telecommunications). Service Access - the ability for the network to provide user access to features and to accept user service requests specifying the type of bearer services or supplementary service that the users want to receive from the PCS network (Federal Standard 1037C: Glossary of Telecommunications). Service Integrity - the degree to which a service is provided without excessive impairment, once obtained (ATIS Telecom Glossary, 2000). Service Provider - a company, organization, administration, business, etc., that sells, administers, maintains, charges for, etc., a telecommunications service. The service provider may or may not be the provider of the network (ANSI T1.620-1991). Service Retainability - the ability of the user to keep a service, once it has been accessed, under given conditions for a requested period of time (Ericsson, TEMS Investigation GSM manual). Session - relationship between two entities existing over a period for a specific purpose and having characteristic data and states (University of the Witwatersrand, Johannesburg, Glossary of Telecommunications). Terminal - a device capable of sending, receiving, or sending and receiving information over a communications channel (ATIS Telecom Glossary, 2000). Throughput - the number of bits, characters, or blocks passing through a data communication system, or portion of that system. Note 1: Throughput may vary greatly from its theoretical maximum. Note 2: Throughput is expressed in data units per period of time (ATIS Telecom Glossary, 2000). Traffic - the information moved over a communication channel (ATIS Telecom Glossary, 2000). Traffic load - the total traffic carried by a trunk or trunk group during a specified time interval (ATIS Telecom Glossary, 2000). User - a person, organization, or other entity (including a computer or computer system), that employs the services provided by a telecommunication system, or by an information processing system, for transfer of information. A user functions as a source or final destination of user information, or both (Federal Standard 1037C: Glossary of Telecommunications).



INTRODUCTION As the Internet is becoming more and more mobile, it is important for cellular network operators to improve the performance of their networks in order to achieve a better performance of Internet applications that mobile users can access. The reason to perform these improvements is the growing user demand on the quality of service (QoS) of these applications. According to ITU E.800 Recommendation [6] and ETSI TS 102 250-2 V1.2.1 [12], end-user QoS of a particular service consists of three main aspects: Service Access (also referred as Service Accessibility), Service Retainability and Service Integrity. To say it in more simple words- it is what a user expects from the performance of this service, e.g. how easily it can be accessed, how long does it take for a connection to be established, will the received data be defected or not, and so on. Technically these expectations can be expressed through a set of QoS parameters, also referred as Performance Indicators (e.g. unavailability, session setup time, mean data rate, packet loss, round trip time, etc). In GSM 2,5G networks packet data is handled by General Packet Radio Services (GPRS). The problem is that in the beginning of GSM era (early 1990-s) the target of a digital cellular communication system was to provide circuit-switched voice communication with a better quality than it was offered by GSM’s “ancestor”, NMT analogue systems. Hence, the priority of circuit-switched speech has been set higher than the priority of packet data services and there were no certain QoS requirements set for GPRS, so, at the very beginning the service had been offered at “best effort”, without any guarantees to end user. However, GPRS became more and more popular among GSM subscribers, and this fact resulted in setting certain QoS requirements for packet data services. But the priority of packet data is still lower than of conventional circuit-switched voice communication in GSM networks today. From the other hand, increase of GPRS users will increase the demand on performance/QoS of GPRS-based Internet applications that nowadays become more and more mobile-oriented. This is a challenge for GSM operators to improve their GPRS performance, so that it could meet end-user QoS requirements. But there are certain obstacles that operators will face when they try to make these improvements. One of these obstacles is the access environment- GSM radio network. Radio channel conditions (e.g. interference) and bandwidth availability are the factors that have their impact on network performance and, as a result, on end-user quality of service of GPRS applications. GSM network operators have to minimize the impact of radio network-related factors to a certain extent, so that GPRS-based applications could be offered with a guaranteed quality. The purpose of this work is to define the impact of these factors on the performance/QoS of GPRS-based applications in a particular GSM network. This work consists of two parts. The first part (chapters 1, 2 and 3) contains theoretical background. Chapter 1 gives a general overview of GPRS system, including network nodes, interfaces and basic mobility management procedures. Chapter 2 describes general concepts of QoS, including QoS customer-provider formation model (Quality Cycle), QoS levels and QoS aspects, defined by ITU Recommendation E800, with the set of QoS parameters for GPRS applications. This chapter also includes some requirements for GPRS applications regarding QoS (e.g. service precedence and delay). Chapter 3 gives an overview of radio network factors that affect GPRS performance. The factors described in this chapter are:

• Bandwidth availability, which depends on the configuration of packet data channels and current traffic load in a cell,

• Radio link quality, which is defined by the level of co-channel interference expressed through carrier-to-interference ratio (C/I). In GPRS there are 4 channel coding schemes, and they can be used in different interference conditions. As these conditions are changing, an algorithm of switching between coding schemes (Link Adaptation) can be



implemented to provide more interference-tolerant packet data transmission over the air interface.

The second part (chapters 4 and 5) is a research part and contains the information about tested system and tests’ results with their analysis. Chapter 4 describes the tested system- EMT GSM network with a short overview of its GPRS structure. There is also a description of the test system- a set of hardware and software tools, with Ericsson TEMS™ Investigation GSM software as the most used tool. In the end of the chapter there are three test scenarios of packet data channel configuration versus cell traffic load. The results of the tests, conducted according to these scenarios, as well as the comparison and analysis of these results, are given in Chapter 5. This work also contains several appendixes that contain

• Information about the methods of measuring mean data rate • Information about GPRS coding schemes and link adaptation algorithms • Information about retransmission with Selective Automatic Repeat Request (ARQ) • Description of EDGE



1. GPRS OVERVIEW This chapter gives a general overview of General Packet Radio Service (GPRS) in a Global System for Mobile communications (GSM) network. The following aspects are included:

• GPRS architecture and interfaces: new nodes, updates to existing GSM network nodes (see 1.2.1-1.2.2),

• GPRS interfaces and protocols (see 1.2.3-1.2.4), • GPRS Mobility management: main mobility management procedures (see 1.3), • GPRS applications: a list of applications that can use GPRS as data bearer (see 1.4).

1.1 Introduction to GPRS General Packet Radio Service (GPRS) provides packed data services to the GSM and WCDMA (Wideband Code Division Multiple Access) systems [1]. It provides a basic solution for Internet Protocol (IP) communication between Mobile Stations (MS) and the Internet or corporate Local Area Networks (LAN).

GPRS provides:

• Efficient transport of packets in the cellular network • Efficient use of scarce radio resources • Flexible service, with prepaid or postpaid charging based on content, volume, or

session duration. • Fast setup and access time • Simultaneous circuit-switched and packet-switched services, which means

coexistence without disturbance • Connectivity to other external Packet Data Networks (PDNs), using IP

GPRS data transfer is IP based. A message consisting of large quantities of data is divided into several packets. When these packets reach the destination, they are stored in data buffers and reassembled to form the original message. The packet data transmission is thus carried out on an end-to-end basis, including the radio interface.

GPRS is another step in the evolution of mobile data services. This is the next step from High Speed Circuit-Switched Data (HSCSD). The evolution of mobile data services by the year 2002 is given in Figure 1.1.



Figure 1.1 The evolution of Mobile Data Services (by the year 2002) [2] Nowadays, an enhancement to General Packet Radio Services- EDGE (Enhanced Data Rates for Global Evolution or Enhanced Data Rates for GSM Evolution) is being implemented. EDGE is the next step in the evolution of GSM. Its objective is to increase data transmission rates and spectrum efficiency and to facilitate new applications and increased capacity for mobile use [3]. EDGE is a method to increase the data rates on the radio link for GSM. Basically, EDGE only introduces a new modulation technique and new channel coding schemes that can be used to transmit both packet-switched and circuit-switched voice and data services. EDGE is therefore an add-on to GPRS and cannot work alone. By adding the new modulation and coding schemes (for coding schemes see 3.2) to GPRS and by making adjustments to the radio link protocols, it offers significantly higher throughput and capacity.

The scope of present work covers “regular” GPRS only. EDGE is, therefore, outside of the scope of this work. A more detailed description of EDGE is given in Appendix E.

1.2 GPRS architecture and interfaces GPRS is not a new system, but a packet-switched add-on to existing GSM circuit-switched network (see Figure 1.2).

Figure 1.2 GSM network with GPRS functionality



The introduction of GPRS results in:

• Adding new nodes to existing GSM network (SGSN, GGSN); • Hardware and/or software upgrade of existing GSM network nodes (e.g. adding PCU to

BSC, software upgrade to BSC and MSC) [1]. The structure of GSM network with GPRS add-ons and interfaces is shown in Figure 1.3. Existing nodes that require software upgrade, hardware and software upgrade as well as new nodes distinct in colours on this scheme.

Figure 1.3 GPRS architecture and interfaces

1.2.1 GSM network nodes

As you can see in the Figure 1.3, the main GSM network elements that GPRS introduction has impact on, are:

• Mobile Station (MS) • Base Transceiver Station (BTS) • Base Station Controller (BSC) • Mobile Switching Center (MSC) • Visitor Location Register (VLR) • Home Location Register (HLR)

1.2.1.1 Mobile Station

Mobile Station (MS) is a combination of a Mobile Terminal (MT), which can be a cellular telephone, and a Terminal Equipment (TE), which can be a computer connected to the MT. It can be concluded from the context, which parts would relate to the MT or the TE parts. The MT and



TE parts can be integrated in one piece of equipment [1]. The scope of GPRS does not cover the communication between the MT and the TE.

The GSM MSs can operate in one of the following modes of operation (for GPRS) [1]:

• Class A mode of operation: The MS is attached to both GPRS and other cellular network services and can operate both services simultaneously;

• Class B mode of operation: The MS is attached to both GPRS and other cellular network services but can operate only one set of services at a time;

• Class C mode of operation: The MS is exclusively attached to GPRS. There are several types of mobile phones:

• Type 2+1: 2 downlink channels and 1 uplink data transmission channel • Type 3+1: 3 downlink channels and 1 uplink data transmission channel • Type 4+1: 4 downlink channels and 1 uplink data transmission channel

The supported data transmission speed per channel is 13,4 kb/s [4].

1.2.1.2 Base Transceiver Station Base Transceiver Station (BTS) is radio equipment, which transmits and receives information over the radio interface to enable the BSC to communicate with MSs in the BSC area. A group of BTSs is controlled by one Base Station Controller (BSC). For GPRS implementation the software upgrade must be done for BTS [1]. However, for the implementation of EDGE a hardware upgrade (an EDGE-capable transceiver) for a BTS is required [3].

1.2.1.3 Base Station Controller

The Base Station Controller (BSC) has the functions to set up, supervise, and disconnect circuit-switched calls and packet-switched data sessions. It has a high capacity switch that provides functions such as handover, cell configuration data, and channel assignment. One or several BSCs are served by one Mobile Switching Center (MSC) for call handling, and a number of BSCs are served by one SGSN for packet data transmission. A BSC can, however, be connected to only one SGSN.

For GPRS implementation the hardware and software upgrades have been done for BSC. The new hardware unit in BSC is Packet Control Unit (PCU). Its function is to separate packed-switched data from circuit-switched voice traffic [1].

1.2.1.4 Mobile Switching Center

The Mobile Switching Center (MSC) handles circuit-switched transactions with the MSs located in a geographical area designated as the MSC area. It is responsible for the set-up, routing, control, and termination of the call, for management of inter-MSC handover (the procedure of updating user location information while the user is moving from one cell to another), for supplementary services, and for collecting charging information.

For GPRS implementation the software upgrade must be done for MSC [1].



1.2.1.5 Visitor Location Register

Visitor Location Register (VLR) is a database containing temporary information about all the MSs currently located in the area served by one MSC. VLR is usually implemented in the same node with MSC [1].

For GPRS implementation the software upgrade must be done for VLR [1].

1.2.1.6 Home Location Register

Home Location Register (HLR) is a database that holds subscription information for all MSs subscribing to the GSM cellular network. The HLR stores information for circuit-switched and packet-switched communication, for example, the location of the MS, supplementary services, authentication parameters, and whether or not packet communication is allowed. For the packet-switched part of the network, subscription information is fetched from the SGSN (see 1.2.2.1), and the HLR is used for the authentication procedure of the MSs. In addition, the HLR stores the subscription information regarding MS-terminated SMS messages and the option of SMS message transfer through the SGSN (see 1.2.2.2).

For GPRS implementation the software upgrade must be done for HLR [1].

1.2.2 GPRS nodes The two main nodes are:

• Serving GPRS support node (SGSN); • Gateway GPRS support node (GGSN).

1.2.2.1 Serving GPRS support node Serving GPRS Support Node (SGSN) is a primary component of cellular networks that employ GPRS. Via the radio network, the SGSN routes incoming and outgoing IP packets addressed to or from any GPRS subscriber physically located within the geographical area served by that SGSN [5]. Each SGSN provides:

• Ciphering (encryption and decryption) and authentication; • Session management and communication; • Set-up to the mobile subscriber; • Mobility management (support for roaming and handover within and between mobile

networks); • Logical link management to the mobile subscriber connection to other nodes (HLR, MSC,

BSC, Short Message Service Center- SMSC, GGSN); • Collecting charging data for each mobile subscriber, such as the actual use of the radio

network and GPRS network resources.

1.2.2.2 Gateway GPRS support node Gateway GPRS Support Node (GGSN) is also a primary component of cellular networks that employ GPRS. The GGSN serves as the interface to external IP packet networks, accessing external ISP functions such as routers and remote access dial-in user service (RADIUS) servers.



In terms of the external IP network, the GGSN routes the IP addresses of subscribers served by the GPRS network, exchanging routing information with the external network [5]. The GGSN sets up communication with external networks and manages GPRS sessions. It also includes functionality for associating subscribers to the appropriate SGSN. For each mobile subscriber, the GGSN also collects charging data (use of the external data network and use of GPRS network resources).

1.2.3 GPRS interfaces Standards from ETSI and the Third-generation Partnership Project (3GPP) specify several logical interfaces to and from the GSNs. Some of these are described in Table 1.1.

Table 1.1 GPRS interfaces Interface Nodes Used for Gn, Gp GGSN - SGSN • Control signalling for mobility and session management

• Tunnelling of end-user data payloads in the backbone network.

• Gp interface is used when SGSN and GGSN are located in different networks

Gb SGSN - BSS SGSN signalling with the BSCs in GSM or TDMA packet-access networks.

Gr SGSN - HLR MAP signalling to support storage and retrieval of subscriber data between the SGSN and HLR.

Gs GSN - MSC • Providing mobility management to subscribers (location updates, paging)

• Conveying ANSI signalling messages for TDMA Gi GGSN -

external data networks

• Transportation of end-user IP data between the mobile network and external IP networks,

• GGSN control signalling with ISP servers located in IP networks (including end-user authentication and IP address allocation via RADIUS [5]).

Iu SGSN - BSC • Carrying IP traffic between the core network and the radio network.

• SGSN control signalling between the radio network and the core network.

Gf SGSN - EIR MAP signalling to support identity-check procedures between the SGSN and EIR servers when a user is attaching

Gd SGSN - SMSC MAP signalling to support the SMS service over packet-switched radio channels

Gm SGSN – PTM GGSN - PTM

• Signalling between the PTM-SC, the SGSN and the GGSN • Carrying messages between these nodes after a request

has been made by a PTM server application to send data to a group with or without geographical filtering.

• Gm interface is currently being specified by the 3GPP Gc GGSN – HLR An optional interface that allows the GGSN access to customer

location information



1.2.4 GPRS protocols SNDCP

• Sub-Network Dependent Convergence Protocol • Compression, Segmentation and Multiplexing of network layer messages towards a

single virtual connection [4] LLC

• Logical Link Control • Ensures the reliable transfer of user data across a wireless network (MS and SGSN)

BSSGP

• Base Station System GPRS Protocol • Routing and QoS information for the BSS

GTP (GPRS Tunnelling Protocol)

• Tunnels the protocol data units through the IP backbone by adding routing information [4] GPRS protocol stack is shown in Figure 1.4:

Figure 1.4 GPRS protocol stack [1]

1.3 GPRS mobility management IN GPRS MS can be in 3 states:

• Idle - no MS location information is kept by the network, no PDP (Packet Data Protocol, see 1.4.2) context is active. After a successful MS Attach procedure, the state is changed to ready [1].

• Ready - The location is known at cell level, an active PDP context to an access point name is established

• Standby - The location is known at the Routing Area Identifier (RAI) level, the PDP context is active

Figure 1.5 shows how these states are being changed.



Figure 1.5 GPRS MS states

1.3.1 MS Attach and detach

Before sending IP packets, the MS must perform an MS Attach procedure. The MS Attach procedure makes the network aware of the MS being present in the network. The procedure is initiated by the MS toward the SGSN. An attached MS is able to perform a Packet Data Protocol (PDP) Context Activation procedure (see 1.3.2.1) to start packet-switched data transfer [1].

The attach procedure is shown in Figure 1.6.

Figure 1.6 MS Attach

Upon completing the Attach procedure, the network is able to track the MS (via subsequent location updates) and is aware of the services and networks that the user has access to [1].



At this point the user is not able to send or receive data.

The MS can be explicitly detached from GPRS by the MS or the network (SGSN/HLR). The MS can also be implicitly detached by the network. The MS-Initiated Detach procedure informs the network that an MS wants to detach. After detachment the SGSN keeps the MM context information. This way, the SGSN does not need to contact the HLR when the MS reattaches, provided that it is still within the same SGSN service area. The Network-Initiated Detach procedure is initiated by the SGSN (explicitly or implicitly) or the HLR. The active PDP contexts are deleted during the Network-Initiated Detach procedure [1].

1.3.2 PDP Context activation/deactivation

1.3.2.1 PDP Context activation

The MS-initiated PDP context activation procedure is used to activate a PDP context in order to establish a virtual data channel through the GPRS network between the MS and a PDN. The MS must be attached to the GPRS network in order to activate a PDP context [1].

A PDP context includes:

• Type of the PDP network • Address of the terminal • IP address of SGSN • Access Point Node (APN) logical name (e.g. wap.emt.ee) • QoS

When the MS requests PDP context activation, the SGSN authorizes the MS and checks the available resources in the GGSN with respect to the load situation (this procedure is called Admission Control) [1]. If the number of active PDP contexts in the SGSN exceeds a predefined limit, the activation is rejected. The SGSN decides which QoS to offer the MS based on information in the subscription from the HLR, received from the GGSN, and SGSN. Also, after Access Point Name (APN) selection, the SGSN interworks with the GGSN for IP address allocation and authorization with PDNs. The APN is a logical name referring to the PDN and to a service to which the subscriber wishes to connect. The APN specifies, among other things, how the IP address should be allocated. The MS will use the allocated IP address throughout the whole session [1].

After a successful PDP context activation, the subscriber is able to communicate with PDNs using the activated PDP context.

PDP Context Activation procedure is shown in Figure 1.7.



Figure 1.7 PDP Context Activation procedure

1.3.2.2 PDP Context Deactivation PDP context deactivation can be initiated either by MS or by network nodes (SGSN or GGSN) [1]. An example of MS-initiated PDP Context deactivation procedure is shown in Figure 1.8.

Figure 1.8 MS-initiated PDP Context deactivation procedure

1.3.3 Data transfer After attaching itself to GPRS (see 1.3.1) and activating a PDP context (see 1.3.2), the MS may start a data transfer session. The MS may send end-user packets on the uplink, as well as to receive end-user packets on the downlink. The packets are routed by the SGSN and GGSN to the correct addresses [1].

1.3.4 Cell Reselection In GPRS, the mobile terminal selects the serving cell [31]. This is different from the basic circuit-switched GSM data service where the network controls the transfer of on-going data calls between cells. The cell reselection process causes a delay, and sometimes packet losses in active data flows. The total delay consists of the radio channel access delay in the Base BSS, and the delay caused by mobility management procedures in the SGSN [1].



Changing between cells that belong to the same BSC can be typically done within the BSS without involving the SGSN, which reduces the delay. Some events in the network may cause cell reselection to be aborted and later restarted which significantly increases the delay. According to the GPRS specifications, a cell reselection should be completed within a few seconds. However, in a live GPRS network it can take any time from a few to a few tens of seconds. The frequency of cell reselections is to a large extent determined by the speed of a user’s movement and the size of cells [1].

1.4 GPRS applications The following GPRS-based applications are available for an end user:

• Information services as text or graphics - e.g. share prices, sports scores, weather reports, news headlines, flight information

• M-commerce - not all content is delivered via the mobile network:

Vending machine, juke box, car wash, and road toll services; Web shopping; Tickets for transport (e.g. trains); Gaming and gambling; Banking

• Location Based Services - personalised information based on the location they are in (maps, hotel information, local Thai restaurants)

• Corporate email – allowing mobile employees to access their internal email system from their office LAN

• LAN applications – mobile employees can access any applications normally available on their PC at the office

• Internet email – by linking Internet email with an alert mechanism such as SMS or GPRS, users can be notified when a new email is received

• Mobile Office - high speed automatic routing of voice and data calls via fixed or mobile links to/from the mobile user

• Web browsing (depending on handset capabilities) • WAP browsing • MMS - Multimedia Messaging Service (MMS) makes it possible for mobile users to send

multimedia messages from MMS enabled handsets to other mobile users and to e-mail users. MMS messages consist of one or more media elements (text, picture, photo, animation, speech, audio, etc), which can be combined and synchronized as to context and time.

• File transfer – downloading sizeable documents and software • SMS – support of transfer of MS-originated and MS-terminated SMS messages between

the MS and the SMSC over the GPRS network.

The most commonly used GPRS-based applications nowadays are WAP, Web browsing, MMS and email.



2. QUALITY OF SERVICE GENERAL CONCEPTS In this chapter general concepts of Quality of Service (QoS) are given. They include Quality Cycle- the mechanism of QoS formation (see 2.2), different QoS levels that a service is offered at, depending on subscribers’ needs (see 2.3) and QoS aspects given in a three-layer model, according to ITU E.800 Recommendation. With regards to the scope of this work, QoS parameters (see 2.4.2) are given particularly for GPRS applications. There are, as well, some requirements and service classes for GPRS presented (see 2.5).

2.1 Definitions of QoS There is no certain single definition of Quality of Service. Different sources give different definitions of QoS. Here is a list of these definitions:

• The collective effect of service performance, which determines the degree of satisfaction of a user for this service [6].

• A set of those quantitative and qualitative characteristics of a distributed multimedia

system, which are necessary in order to achieve the required functionality of an application [7].

• The set of parameters and procedures associated with a service or user, indicating some

of the capabilities and constraints related to the delivery of the service to the user [8].

• A measure of the extent to which user requirements are sustained by the network [10]. Summarizing the definitions listed above we can state the following: the goal QoS is to achieve a certain level of an operator’s network performance to meet the user requirements of a particular service.

2.2 Quality Cycle. Offered versus Delivered QoS Considering the goals and the achievements for each party (customer and provider) involved in the service, one may devise four complementary viewpoints: customers’ QoS requirements / perception, and service provider QoS offering / achievement. Customer requirements are important when creating a QoS test plan to be used when estimating the QoS delivered by a service provider. These requirements are usually expressed in non-technical language and focus on user-perceived effects, rather than their causes within the network. The delivered QoS is expressed in values assigned to QoS indicators (also referred as performance indicators), which are used for tracking performance and directing optimization [9]. The process of QoS negotiation is called Quality Cycle [10]. Figure 2.1 describes this process.



Figure 2.1 The quality cycle [10] QoS Offered:

• No agreement in the telecom industry on how to specify QoS parameters; • Usually in the language of the service provider; • Different services should have their own unique performance parameters; • Number of parameters varies from SP (Service Provider) to SP [10].

QoS Delivered:

• It should measure actual (not the estimated) service delivered. • The measurement should be on an end-to-end basis. • Customer should have tools available to verify service quality [10].

This work is focused mostly on Delivered Quality of Service and its indicators that can be measured on an end-to-end basis by conducting active tests with special software and user terminal equipment, combined into one measurement system (see 4.2).

2.3 QoS levels QoS provides the user/system operator with the ability to provide differentiated levels of service based upon the subscribers application needs. The needs of applications can be used to classify the applications into groups with similar needs. Each group has an associated profile that describes the bounds within service must be provided. The fundamental idea behind QoS is to enable network operators to establish different levels of service pertaining to the end users traffic throughput, traffic loss, and response time and to the charge the end user for the actual level of service provided [11]. Although the many markings in some QoS technologies allow more granular treatment, most service providers offer five levels of service to their customers [11]:

• Premium. Premium level services guarantee delay, jitter (delay variation), and packet loss. This level of service is normally used for real-time traffic (voice, video, and time sensitive data like streaming stock quotes);

• Gold. A gold level of service is protected from dropping, but usually does not have the delay and delay variation (see 2.5.2) guarantees that the premium level of service has. Gold service levels are normally used to transport important data traffic like Enterprise Resource Planning (ERP) applications [11] and other mission critical data traffic.



• Silver. The silver service level is used to transport data that is important but not mission critical to the organization. An example of this might be time reporting and accounting.

• Bronze. The bronze service level is used to carry data traffic that has been identified and should be given better treatment than best effort traffic. An example of this service level might be human resource intranet web resources.

• Best effort. The last and final service level is that of best effort. While the Internet was built on best effort services, this model can break down in enterprise organizations that tax their network infrastructure [11].

2.4 QoS aspects and parameters The QoS aspects and parameters, according to ITU-T Recommendation E.800, are presented here in a three-layer model (see Figure 2.2).

Figure 2.2 A three-layer model of QoS aspects and parameters [12] The first layer is the Network Access, the basic requirement for all the other QoS aspects and QoS parameters. The outcome of this layer is the QoS parameter Network Accessibility. The second layer contains the other three QoS aspects:

• Service Access. ITU E.800 also refers to the term „Service Accessibility“ (see 2.4.1.1), • Service Retainability (see 2.4.1.2), • Service Integrity (see 2.4.1.3).



The different services are located in the third layer. Their outcomes are the QoS parameters [12]. For every QoS aspect of a particular service there is a set of main QoS parameters (ITU E.800 also refers to the term „Key Performance Indicators“, KPI-s). For every main QoS parameter (KPI) there is a set of more simple measures (ITU E.800 refers to the term „Performance Indicators“, PI-s). All parameters have been designed, according to the ITU Recommendation E.800, for performance measurements, that is, to measure service accessibility, reliability and integrity. As the scope of this work covers Quality of Service of GPRS-based applications, this section is focused only on corresponding QoS aspects (see 2.4.1) and parameters (see 2.4.2), defined for GPRS.

2.4.1 GPRS QoS aspects

2.4.1.1 Service Accessibility Service Accessibility is the ability of the user to obtain a service within specified tolerances and under other given conditions [13]. In order for a service to be accessible in a GPRS network, the user must be able to execute a chain of operations:

• Accessing GPRS as such, i.e. performing a GPRS attach and a PDP context activation, • Within an active PDP context, accessing an IP service.

Examples of Service Accessibility parameters for GPRS-based services are:

• GPRS Accessibility Failure Ratio: the percentage of failed attempts to access a GPRS service (see 2.4.2.1),

• GPRS End-to-end Access Time: the time required to set up the GPRS connection, before the data transfer can begin (see 2.4.2.2).

2.4.1.2 Service Retainability Service Retainability is the ability of the user to keep a service, once it has been accessed, under given conditions for a requested period of time. Retainability of a service or session also implies that the user does not have to perform any manual operations that are not necessary under stable network conditions, such as manual re-activation of the PDP context [13]. An example of Service Retainability parameters for GPRS-based services is GPRS Data Transfer Cut-off Ratio (see 2.4.2.4).

2.4.1.3 Service Integrity Service Integrity indicates the degree to which a service is maintained without major disturbances once it has been accessed. Integrity parameters show the performance of successful service attempts. Even if a service was accessed successfully, the user's perception of the performance may vary greatly, from very good to unacceptably bad [13]. An example of Service Integrity parameters for GPRS-based services is GPRS Mean User Data Rate (see 2.4.2.3).



2.4.2 GPRS QoS parameters A set of QoS parameters (further referred as key performance indicators, as it is stated in [6], [12] and [13]) is given in this section. All of them are defined by ITU E.800 recommendation. In Figure 2.3 you can see a diagram of GPRS Session with relevant performance indicators for all supported data services except WAP and MMS. A diagram of GPRS data session for WAP is given in Figure 2.4 [13]. MMS QoS is outside the scope of this work. Main QoS parameters are referred as Key Performance Indicators (KPIs), while sub-parameters, that form a main parameter, are referred as Performance Indicators (PIs) [6], [12], [13]. These parameters are listed in 2.4.2.1-2.4.2.5. The definitions and equations of these parameters are given according to [13].

Figure 2.3 KPIs, PIs, and timeouts for all supported data services except WAP and MMS [13]



Figure 2.4 KPIs, PIs, and timeouts for WAP over GPRS [13]

2.4.2.1 GPRS Accessibility Failure Ratio Denotes the probability that a subscriber cannot access the service successfully [13]. This means that the data transfer of the content could not be started due to a failure that occurred somewhere in the service access chain (GPRS Availability Attach PDP Context Activation Service Access, for WAP there is also WAP Activation procedure after Service Access). For all services except WAP and MMS this parameter is calculated as:

[ ] ( ) 100)1()1()1()1(1% 14131211 ×−×−×−×−−=− PIPIPIPIRatioityFailureAccessibil WAPnon

Equation 2.1

For WAP over GPRS it is calculated as:

[ ] ( ) 100)1()1()1()1()1(1% 1514131211 ×−×−×−×−×−−= PIPIPIPIPIRatioityFailureAccessibil WAP

Equation 2.2 Where PI11 ... PI15 are the following performance indicators:



PI11: GPRS Unavailability. Denotes the probability that GPRS is not active in the cell currently used by the customer (applies only in GSM-covered areas):

10013Re

13Re×=

formationadSystemInptsToTotalAttemformationadSystemInAttemptsTosuccessfulNumberOfUnlityUnavailabi

Equation 2.3

System Information 13 is GPRS specific signalling, that is read once per measurement cycle, at the beginning of the cycle [12]. PI12: GPRS Attach Failure Ratio. Denotes the probability that a subscriber cannot attach to the GPRS network:

[ ] 100Re

% ×=sachAttemptquestedAttrOfTotalNumbe

mptsAttachAttesuccessfulNumberOfUnureRatioAttachFail

Equation 2.4 PI13: GPRS PDP Context Failure Ratio. Denotes the probability that a subscriber cannot activate a PDP context:

[ ] 100Re

% ×=emptsivationAttContextActquestedPDPrOfTotalNumbe

AttemptsAvtivationPDPContextsuccessfulNumberOfUnioFailureRatPDPContext

Equation 2.5

PI14: GPRS Service Access Failure Ratio. Denotes the probability that a subscriber cannot access the service successfully, meaning that the data transfer of the content could not be started:

[ ] 100Re

% ×=sessAttemptquestedAccrOfTotalNumbesessAttemptServiceAccsuccessfulNumberOfUnRatioessFailureServiceAcc

Equation 2.6

PI15: GPRS WAP Activation Failure Ratio. Denotes the probability that the subscriber cannot activate the WAP session:

[ ] 100Re

% ×=AttemptsActivationquestedWAPrOfTotalNumbe

sionAttemptWAPActivatsuccessfulNumberOfUnRatioionFailureWAPActivat

Equation 2.7



2.4.2.2 GPRS End-to-end Access Time GPRS End-to-end Access Time denotes the length (in seconds) of the time period taken to access a service successfully, i.e. to complete the chain GPRS Availability Attach PDP Context Activation Service Access (for WAP over GPRS there is also WAP Activation procedure after Service Access). For all services except WAP and MMS this parameter is calculated as:

[ ] 232221 PIPIPIscessTimeEndToEndAc WAPnon ++=−

Equation 2.8

For WAP over GPRS, this parameter is calculated as:

[ ] 24232221 PIPIPIPIscessTimeEndToEndAc WAPnon +++=−

Equation 2.9

Where PI21 ... PI24 are the following performance indicators:

• PI21: GPRS Attach Setup Time. Denotes the length (in seconds) of the time period taken to attach to the GPRS network.

• PI22: GPRS PDP Context Activation Time. Denotes the length (in seconds) of the time

period taken to activate a PDP context.

• PI23: GPRS Service Access Time. Denotes the length (in seconds) of the time period taken to access a data service successfully.

• PI24: GPRS WAP Activation Time. Denotes the length (in seconds) of the time period

taken to activate the WAP session.

2.4.2.3 GPRS Mean User Data Rate GPRS Mean User Data Rate denotes the average user data rate (application throughput):

[ ] [ ][ ] [ ]stTriggerTimeOfStarsTriggerTimeOfStopkilobytesFileSize

skbittaRateMeanUserDa

−×

=8

Equation 2.10

For this parameter there are two methods of defining a triggering point: one preferring the payload throughput philosophy and the other preferring the transaction throughput philosophy:

• Method A defines trigger points, which are as independent as possible from the service used, therefore representing a more generic view (payload throughput).

• Method B defines trigger points on application layer, therefore representing a more service-oriented view (transaction throughput).

An example of the different trigger points defined for each set is given in Appendix A.



2.4.2.4 GPRS Data Transfer Cut-off Ratio Denotes the probability that a data transfer cannot be completed, due to a network timeout or to the transfer being aborted:

[ ] 100% ×=nsfersrOfDataTraTotalNumbe

staTransfercompleteDaNumberOfIntioerCutOffRaDataTransf

Equation 2.11

One possible reason for data transfer cut-off is PDP context cut-off. The latter occurrence is tracked by a parameter GPRS PDP Context Cut-Off Ratio. It denotes the probability that a PDP context is deactivated without being initiated intentionally by the user:

[ ] 100% ×=textsatedPDPConfullyActivrOfSuccessTotalNumbeertiatedByUsssesNotIniPContextLoNumberOfPDoCutOffRatiPDPContext

Equation 2.12

2.4.2.5 Service Round Trip Time The round trip time is the time required for a packet to travel from a source to a destination and back. It is used to measure the delay on a network at a given time. Round trip time is measured in milliseconds and calculated as:

[ ] [ ] [ ]mstmstmsimeRoundTripT PacketSentceivedPacket −= Re

Equation 2.13

2.5 QoS requirements for GPRS-based applications QoS Requirements for GPRS are stored in QoS Profile [14]. QoS Profile describes applications characteristics and QoS requirements with regards to 4 categories:

• Service precedence – 3 classes (high, normal, low) • Reliability (see 2.5.1) – probability of loss, duplication, missequencing, corruption of

packets - 3 classes • Delay (see 2.5.2) – 4 classes • Throughput – maximum and mean bit rate (see the description in 2.4.2.3)

QoS profile is included in PDP context. Negotiation is managed through PDP procedures (activation, modification and deactivation). QoS profile can be negotiated for each session [14].

2.5.1 Reliability Currently, the support of differentiated QoS (in this case traffic is grouped in service classes that are served differently by the network, one service class may be treated better than another) is minimal [15]. However, GPRS does make it possible to ensure the integrity of received data



through the implementation of two reliable modes of operation: RLC Acknowledged and LLC Acknowledged (see 3.2).

RLC acknowledged mode is used by default to ensure that the data received by/from the MS is without error.

LLC acknowledged mode is an optional feature that may be provided. This protocol ensures that all LLC frames are received without error. However, use of this protocol has an impact on throughput since the correct receipt of all LLC frames has to be acknowledged [15].

2.5.2 Delay (latency and jitter)

Latency is the time taken for data packets to pass through the GPRS bearer, normally measured as a round trip time (see 2.4.2.5). Jitter is the variability in round trip time [15].

In GPRS there are a number of factors contributing to the overall latency. These include:

• Mobile Station (MS), • Radio resource procedures, • Effective data throughput • GPRS core network nodes.

Mobile Station (MS) delay is the time taken by the MS to process an IP datagram and request radio resource. This includes the delay from the PC to MS, and the MS processing time.

Radio resource procedures are the major source of delay in GPRS. In order for the MS to be capable of sending or receiving data, radio resource known as a Temporary Block Flow (TBF) must be made available to the user [16]. If a TBF is currently active then the MS may use it hence minimizing the delay. However, if no TBF is established then the MS and network must exchange signaling messages in an attempt to establish a TBF. The time taken to successfully achieve an active TBF will depend on the availability of radio resources and will be different for the uplink and downlink directions. Once established, the TBF will generally remain active for as long as data is made available to the layer (i.e. for as long as there are LLC frames to transmit) [15].

Effective data throughput (over-the-air delay) is the rate at which user data is physically transmitted between the MS and the SGSN over an active TBF. The delay associated with this throughput is directly related to the size of the IP datagram being sent. Smaller packets cause less delay. The delay is proportionally reduced when multiple timeslots are used. The effective throughput is also dependent on the number of re-transmissions resulting from the hostile radio environment (i.e. the RLC Block Error Rate). The time taken to re-transmit erroneously received information will affect the size of the delay [15].

Core network delay occurs as packets transit through the SGSN and GGSN. These nodes effectively operate as IP routers and as such will have a relatively low impact on the overall latency. However, under high load conditions the transit delay may increase [15].



3. RADIO NETWORK FACTORS THAT AFFECT GPRS QOS

One of the main aspects of QoS/performance of a particular service is a number of factors that have impact on it. In case of GPRS applications, these factors include:

• Radio network factors (bandwidth availability, load, channel quality) • PCU capacity and performance • Performance of GPRS nodes (SGSN and GGSN) • Application server performance (incl. server availability)

The scope of this work covers those factors that are related to radio network as an access environment to the applications. The factors that are related to radio network include:

• Bandwidth availability • Radio channel quality • User mobility (e.g. cell reselections)

The first factor includes such aspects, as Packet Data Channel (PDCH) configuration and traffic load. They are described in 3.1. The second factor is related to channel interference and is described in 3.2. User mobility factor is outside of the scope of this work.



3.1 Bandwidth availability One of the main bottlenecks in cellular networks is bandwidth. This is especially important due to fact, that speech is more privileged service than data. This results in limited amount of bandwidth available for packet data transmission. Also we have to keep in mind user’s mobility (handovers- changing cells).

3.1.1 GPRS channels

3.1.1.1 GPRS physical channels

Figure 3.1GSM physical channels [14]

Figure 3.2 shows the principle of packet data channels allocation for GPRS users.

Figure 3.2 Allocation of packet data channels in GSM network [17]

3.1.1.2 GPRS logical channels

In GSM there are traffic and signalling/control logical channels that are mapped on physical channels. The same situation is about GPRS. Table 3.1 lists GPRS logical channels.



Table 3.1 GPRS logical channels [14]

Group Channel Function Packet data traffic channels PDTCH Data traffic

Packet broadcast control channels PBCCH Broadcast control PRACH Random access PAGCH Access grant PPCH Paging

Packet common control channels

PNCH Notification PACCH Associated control Packet dedicated control channels PTCCH Timing advance control

Logical channels have to be mapped on physical channels, so we need to use bandwidth not only for transmission/reception of packet data, but also for control and signalling.

3.1.2 PDCH configuration

The main problem is the configuration of a particular GSM cell with regards to Packet Data Channel (PDCH) allocation. As it has been mentioned before, voice communication has priority over data communication in a GSM network. An operator has to choose how to allocate packet data channels in a particular cell, considering this priority. There are 3 types of PDCH configuration in a cell: dynamic (On demand), static (Dedicated) and combined [17].

3.1.2.1 Dynamic PDCH configuration

In case of dynamic PDCH configuration every traffic channel that is not used for speech at the moment can be used for packet data. But the problem is in the priority that circuit-switeched speech communication has over packet-switched data communication in GSM. This means, that in case of intensive speech traffic in a cell, it may occur, that all the GSM traffic channels used for GPRS will be overtaken for needs of voice communication. It also may occur, than a subscriber cannot even initiate a GPRS session because all the traffic channels are occupied by voice communication. In both cases it may result into channel assignment reject and expiry of Channel Request Timer. This timer is defined by GSM 0408 specification for several generations of mobiles [27] and is used to control how long the MS should wait with sending a new channel request after having received immediate assign reject. Channel assignment reject and channel request timer expiry result either into GPRS Unavailability (see 2.4.2.1) or a data transfer/session cut-off (see 2.4.2.4). Therefore, this configuration is not „GPRS-friendly” in case of intensive voice traffic in a cell.

3.1.2.2 Static PDCH configuration

In case of static PDCH configuration we have a fixed number of traffic channels assigned for packet data, but, from the other hand, there are no channels on demand. This means, that if all the data channels are occupied, there is no chance for a new subscriber to set up a GPRS session before any of occupied channels is free [17].

3.1.2.3 Combined PDCH configuration

Another possibility for a network operator is to combine both static and dynamic PDCH configurations, as it is shown in Figure 3.3. In this case, the number of dedicated channels for packet data should depend on particular cell characteristics (e.g. capacity, traffic intensity).



Figure 3.3 Combined static and dynamic configuration of PDCH channels in a cell [17]

3.2 Interference Even if there is enough bandwidth to transmit desired amount of packet data over the air interface, we have to keep in mind the current radio channel conditions. It is well known, that one of the major limitations in cellular telephone networks is the so-called co-channel interference. In the case of TDMA networks, such as GSM/GPRS, the co-channel interference is mainly caused by the spectrum allocated for the system being reused multiple times (“frequency reuse”) [18]. The problem may be more or less severe, depending on the reuse factor, but in all cases, a signal received by an MS will contain not only the desired forward channel from the current cell, but also signals originating in more distant cells. Co-channel interference, when not minimized, decreases the Carrier-to-Interference Ratio (CIR, C/I) at the periphery of cells, causing diminished system capacity, more frequent handoffs, and dropped calls. The carrier-to-interference ratio is the ratio of the power in the carrier to the power of the interference signal. The carrier-to-interference ratio is normally expressed in dB [18]. The increase of interference and, correspondingly, decrease of CIR both indicate reduction of radio channel quality. To provide reliable data transmission we need to protect the data we transmit with certain amount of redundancy bits. This, from the other hand, will lead to increase of the total number of bits (information plus redundancy) to transmit and require more time for the transmission of information. Considering this we have to reach the trade-off between reliable and fast transmission with regards to current radio channel quality. This can be reached by channel coding and link adaptation with the possibility of retransmission using selective ARQ.

3.2.1 Channel coding The purpose of channel coding is to protect data packets against errors that may occur during over-the-air transmission. Prior to transmission, data packets are segmented into smaller data blocks across the different layers, with the final logical unit being the Radio Link Control (RLC) Block [19]. More detailed information about data flow segmentation and encoding of GPRS data packets you can find in my other work, dedicated to GPRS coding schemes and link adaptation [20] and in Appendix B.

The RLC block’s data field length will depend on the channel Coding Schemes (CSs) used. Four channel coding schemes, CS1 to CS4, are specified for the GPRS packet data traffic channels [20]. Each scheme has been designed to provide different resilience to propagation errors under unfavorable radio conditions, offering a trade-off between throughput and coding protection. CS1 corresponds to the more robust scheme while CS4 does not use any error correction. Radio Block structures and parameters of four coding schemes are also shown in Appendix B.



3.2.2 Link adaptation Decreasing C/I can lead to the point, where it is no longer possible to provide reliable transmission with current coding scheme. This may result in number of retransmissions and even in loss of data. To protect data packets we can use link adaptation. The objective of link adaptation is to achieve maximum throughput by selecting a suitable coding scheme instantaneously. The choice of coding schemes can be described as a function of the C/I of the channel, i.e. if the C/I is known a coding scheme that maximizes the throughput can be chosen (see Figure 3.4).

Figure 3.4 Throughput versus carrier to interference ratio [21]

Two link adaptation algorithms exist in theory. They are estimated C/I algorithm and Block Error Rate (BLER, see also 3.2.3)-based algorithm. More info about these algorithms is given in [20] and in Appendix C.

3.2.3 Coding schemes for retransmissions. Selective ARQ. When retransmissions occur in the system it is not always reasonable to change to a stronger code (this means- a code with a bigger number of redundancy bits). If current C/I is low, it may still be relevant to keep a weaker coding scheme due to the benefit of a higher bit rate in the transmissions. Retransmissions in themselves, however, are always costly since they require extra signaling. It can therefore be relevant to apply a stronger coding to the retransmissions than to the ordinary transmissions to avoid new retransmissions, even if current C/I is below the threshold for changing schemes [19] (see Figure 3.4). Transmission of radio blocks over the radio interface is controlled by a selective ARQ (Automatic Repeat Request) mechanism [22], whose function consists of error detection and retransmission. In GPRS the transmitter side has a sender window of 64 blocks and the receiver side has a same size sender window accordingly. When the transmitter sends blocks from the sender window, the receiver sends back Temporary Packet acknowledgements to indicate whether the received block is correct or erroneous [22]. The principle of Selective ARQ is described in details in Appendix D.



4. TESTED SYSTEM AND TEST SETUP The goal of this work is to show the impact of radio network factors mentioned in Chapter 3 on the QoS parameters defined by ITU E.800 Recommendation (see 2.4.2) for GPRS applications. To do it, we need to perform a number of tests that contain measurements of these parameters and analysis of fault cases (e.g. GPRS Unavailability or PDP Context Activation Failure; see 2.4.2). This chapter describes the tested system, measurement tools and test scenarios. As it is stated in 2.2, the purpose of the experimental part of this work is to measure Delivered QoS parameters on an end-to-end basis with user equipment, which means, to perform a number of active tests. These active tests were performed in three different cell traffic load and PDCH configuration conditions. The results of the tests are sets of parameters and they are given in Chapter 5. In this chapter there is a short overview of tested system- EMT GSM network with its’ GPRS structure, an overview of tools needed to perform these active tests (Ericsson TEMS™ Investigation GSM toolkit for measurements on Um interface, Agilent acceSS7 toolkit for monitoring on GPRS interfaces and SGSN events logging) and test scenarios for 3 cases of cell traffic load and PDCH configuration.

4.1 A short overview of tested network The tests will be performed in a GSM network, owned by Estonian GSM operator EMT AS. The network is dual-band (900/1800) and covers the territory of the Reupblic of Estonia. An overview of GPRS system in the tested network is given in 4.1.1.

4.1.1 Overview of EMT GPRS structure EMT introduced the GPRS service in spring 2000. The service has become very popular: by now, there are over 20000 subscribers using GPRS. An internal research has shown, that in mean time there are more than 400 active PDP contexts over the entire network. This research has shown, as well, that the most popular GPRS-based service used in EMT network is WAP (about 70% subscribers), followed by HTTP (~ 25%) and MMS (~5%). EDGE functionality (see 1.2.1.2) is now being implememted. At the present moment there is a number of EDGE-capable base stations available in bigger towns (like Tallinn and Tartu). A simplified scheme of EMT GPRS structure is given in Figure 4.1. There is a number of Access Point Nodes (APNs) accessible within EMT network (e.g. internet.emt.ee, wap.emt.ee), as well as some APNs that are physically residing outside of EMT network and are accessed via tunneling protocols. In case of WAP and MMS services this is Generic Routing Encapsulation (GRE) Tunneling (more info about GRE tunneling is given in [23]). For corporate customers IPSec tunneling (more info about IPSec is given in [24]) is used.



Figure 4.1 A simplified view of GPRS structure of the tested network The situation about QoS of GPRS in the network is the following:

• No guaranteed minimum bitrate is specified. • There is a limitation on maximum bitrate for prepaid users. This is to prevent them from

using EDGE, which allows a higher bit rate. • Guarantees concerning accessibility and reliability are specified for core network only. No

special guarantees for the radio network exist at the moment. • No link adaptation with GPRS coding schemes CS1-CS4 is currently in use. In fact, the

network supports only CS1 and CS2 schemes. The last fact is typical not only for this particular operator. It is a normal case for a GSM operator and is caused by a higher priority of circuit-switched voice communication over packet-switched traffic.

4.2. Measurement system This section describes the tools needed to perform active tests of GPRS applications. The most important tool is TEMS™ Investigation GSM software toolkit by Ericsson (see 4.2.1). It is used for measurements on Um interface (for GPRS interfaces see 1.2.3). In addition, two data sources are used: acceSS7 Call Trace for monitoring of GPRS interfaces and SGSN internal logging. The scheme of the measurement system used for active tests is shown in Figure 4.2.



Figure 4.2 Measurement system for GPRS active tests

4.2.1 TEMS™ Investigation GSM – a short description TEMS™ Investigation GSM is an air interface test tool for troubleshooting, verification, optimization, and maintenance of GSM/GPRS/EDGE networks [25]. It measures radio parameters, assesses speech quality, and decodes air interface messages. The data is presented in real time along with information on cell sites and channels, and can be saved for later use. For measurements a special type of handset must be used (in this work it was a Sony Ericsson T610 telephone with a special TEMS™ software version installed). The mobile is connected to terminal equipment (a laptop) through a data cable. Key features of TEMSTM include:

• Measurement of Key Performance Indicators • Identification of interference sources • C/I measurements in real time • Data measurements (HSCSD/CSD/GPRS/RAS/EDGE) • WAP/MMS/SMS testing • Summaries of log file(s) in HTML format • Command sequence tool for automating test procedures

One of the main features is measurement of KPIs. TEMS™ Investigation provides the set of KPIs defined by ITU Recommendation E.800 (see 2.4.2). Typical services that can be used for testing are HTTP, WAP, and PING (to measure Round Trip Time). Measurements can be utilized from the command sequence tool in order to automate repeated procedures [25].



All measurements can be saved to log files, where they are time-stamped. Data from up to four TEMS™ mobiles and a TEMS™ Scanner can be logged simultaneously. Using command sequences, various tests can be automated. Log files can be exported as user-defined tab-delimited text, plain text, Planet (DMS 3.1) or in MapInfo Tab or MIF format [25]. A log file or a set of log files can be summarized in user-definable reports (HTML files). The reports contain charts and statistics for parameters and events of the user’s choice. The collected data is presented on a map, in line charts, and in text format windows, all of which are synchronized. Events and air interface messages are listed separately and can be inspected in detail. The user interface can be freely configured, and users can build their own presentation windows [25]. A screenshot of TEMS™ user interface is shown in Figure 4.3.

Figure 4.3 TEMS™ Investigation GSM user interface

4.2.2 Additional tools: SGSN logging and acceSS7 monitoring. In addition to measurements on Um interface handled by TEMS™ Investigation GSM, there are two more sources of data to be used in active tests. They are:

• acceSS7 monitoring (by Agilent); • SGSN logging (by Ericsson)

Agilent acceSS7 is a distributed OSS (Operation and Support System) that collects and analyzes messages from the SS7 signalling links from a variety of wireless or wireline technologies, including GSM, GPRS, optical, fixed voice, NgN, VoP and IP networks [26]. It is totally switch-



independent, providing a comprehensive, impartial view of what is happening on the network even during fault conditions. By mining and interpreting the signaling data on the SS7 network [26] acceSS7 provides real-time and historical information on the network, calls and services. This data can be used by different functions, including Network Operations, Network Planning & Engineering, Wholesale/Carrier Services, Marketing and Service Operations. In this work acceSS7 monitoring system can be used for detailed analysis of several fault cases registered in TEMS™ logfiles. Gb, Gr, Gn, Gi and Gp interfaces are being monitored (for more info on GPRS interfaces see 1.2.3). SGSN logging can be also used for detailed analysis of several fault cases. Actions, performed by SGSN, (e.g. mobility management procedures, see 1.3), are being monitored by SGSN logging [1].

4.3 Test scenarios There are 3 aspects of test scenarios:

• Radio network conditions • Services to be tested • Parameters to be measured

Radio network conditions that affect QoS parameters of GPRS are listed in Chapter 3. These are:

• Bandwidth (PDCH configuration and traffic) • Interference

To define the impact of the first factor, we need to measure QoS parameters with different PDCH configuration and traffic. Three scenarios are given in 4.3.1. Due to complexity, there will be no special interference scenarios, but in every test TEMS™ Investigation GSM will monitor the interference level and all fault cases will be reviewed for the subject of interference conditions. Two services were chosen for the tests: WAP and HTTP, as the most used GPRS-based services offered by EMT (see 4.1.1). Both services will be tested in every of three scenarios given in 4.3.1. The result of these tests will be in six sets of measured QoS parameters, listed in 2.4.2.

4.3.1 Tested cell: PDCH configuration, cell capacity and load For the active tests a cell located inside of EMT office had been chosen. This cell has total capacity of 14 traffic channels, handled by 2 transceivers. Allocation of a dedicated traffic channel for packet data is enabled in the cell. To perform active tests in different PDCH configuration and cell traffic load conditions we need to:

• Change the configuration • Create extra traffic load

Configuration changes in the cell can be done with a special administrative software tool. Extra traffic can be created in artificial way, by means of setting up a number of circuit-switched speech connections within the cell. As it is hard to achieve high traffic with 14 TCH-s in an indoor cell from one hand and limited amount of user terminals available from the other hand, the capacity of the cell should be reduced by blocking one of two transceiver units serving the cell. As a result, there will be 6 traffic channels available. Therefore, it is enough to have two artificially created circuit-switched connections in the cell to achieve sufficiently high traffic and, from the other hand,



to avoid congestion, because, in this case, there will still 4 traffic channels available for regular communication with dynamic PDCH configuration. With combined PDCH configuration (one dedicated PDCH) there will 3 channels available. According to PDCH configuration types listed in 3.1.2, the following scenarios have been chosen:

• Dynamic PDCH configuration with normal traffic load in the cell. The tests will be performed with full cell capacity of 14 traffic channels and without any extra load artificially created. The purpose of these tests is to show the values of GPRS QoS parameters in normal conditions.

• Dynamic PDCH configuration with high traffic load in the cell. The tests will be performed with cell capacity reduced to 6 traffic channels. Artificial increase of traffic will be provided by setting up 2 circuit-switched connections in the cell. The purpose of these tests is to show, what impact high traffic has on GPRS QoS parameters in case of dynamic PDCH configuration.

• Combined PDCH configuration with high traffic load in the cell. The tests will be performed with cell capacity reduced to 6 traffic channels and with one traffic channel reserved for packet data communication. This channel will be shared by two packet-data users: the first will be the one performing the active test and the second will be an auxiliary test-handset using WAP services. These two will use the channel simultaneously. Artificial increase of traffic will be provided by setting up 2 circuit-switched connections in the cell. This will prevent packet data users from occupying any of the remaining 5 traffic channels for long time. The purpose of these tests is to show, how effectively (first of all- in the contexts of Accessibility and Reliability, see 2.4.1) one dedicated PDCH can be used by two packet data subscribers in case of high load and combined PDCH configuration, to be compared with the previous scenario.

4.3.2 Tested services According to every scenario of those listed in the previous section, two services will be tested: HTTP (Web browsing) and WAP. The active tests will be performed with TEMS™ Investigation GSM, and the results will be given in specially generated reports (see 4.2.1). For HTTP tests, a local search system homepage (http://www.neti.ee) has been chosen. For WAP tests, EMT WAP homepage (wap.emt.ee) has been chosen. For every active test there will be 50 TEMS measurement cycles to measure all the parameters listed in 2.4.2 except round trip time and packet loss. For the last two, 3 PING (Packet Internet Groper) test sessions (every session has a special value of packet size defined) will be performed. For every session, 100 packets will be sent to a server. The server, in this case, is each of the two servers listed in previous paragraph.

4.3.3 Tested parameters For each service of those listed in the previous section a set of key performance indicators will be measured by TEMS™ Investigation GSM. These parameters are measured according to the equations given in 2.4.2. However, in case of GPRS End-to-end Access Time (see 2.4.2.2), mean value of this parameter is calculated for every complete GPRS Access chain (Attach PDP Context Activation Service Access, for WAP also WAP Activation), while mean values of the elements of this chain are measured separately. Therefore, the mean value of GPRS End-to-end Access time is not the same, as if it was calculated as a sum of mean values of these elements, according to Equation 2.8 for HTTP and Equation 2.9 for WAP. In case of PING sessions, round trip time is calculated by TEMS™ Investigation GSM, while Packet Loss Ratio is calculated as following:



[ ] 100% ×=SentrOfPacketsTotalNumbe

stPacketsNumberOfLoRatioPacketLoss

Equation 4.1

The number of lost packets corresponds to the number of PING Timeouts. Every PING Timeout case is registered in a TEMS™ log file. As there are just 50 attempts for every test scenario, the results of these tests do not pretend to be too precise and give a more general view of GPRS services performance in the cases listed in 4.3.1.



5. MEASUREMENTS OF QOS PARAMETERS This chapter shows the results of active tests that have been performed during a 3-week period (November-December 2004), according to the scenarios given in 4.3. For every scenario there is a short overview with the values of most important measured QoS parameters (further in this chapter they are referred as Key Performance Indicators, KPIs). A more detailed view and comparison of measured parameters is given in tables and charts section 5.7.

5.1 KPI measurements: HTTP, normal cell traffic load, dynamic PDCH configuration The test was performed in normal load conditions (see 4.3), with approximate cell traffic load value of 2,2 erlangs.

5.1.1 GPRS Accessibility Failure Ratio All of 50 attach attempts were finished successfully, making therefore Unavailability and Attach Failure Ratio values equal to zero. However, there were two other failures to access GPRS service: one was a failed PDP Context Activation and the other was a failed Service Access attempt. The further analysis of TEMS™ log file has shown that in first case there was an unacceptably high level of interference (below 4,5 dB) for more than 6 seconds, and it resulted in a dropped connection and Channel Request Timeout, so the attempt to start a GPRS session had failed at the stage of PDP Context activation. The analysis of the second case (where GPRS session access attempt has not finished successfully) has shown that the session was broken from the GGSN side with the reason “The address is not valid”. The last message is related to the IP address that was assigned for this session, after successful activation of the PDP Context. This case, however, is obviously not related to radio network factors. It can be used for further researches of other factors’ impact on GPRS key performance indicators. With these two failures, the values of PDP Context Activation Failure Ratio (see Equation 2.5) and GPRS Service Access Failure Ratio (see Equation 2.6) are, correspondingly, 2% and 2,04%. This makes the approximate value of GPRS Accessibility Failure Ratio equal to 4% (calculated by Equation 2.1):

[ ] ( ) %4100)0204,01()02,01()01()01(1% ≈×−×−×−×−−=HTTPRatioityFailureAccessibil

5.1.2 GPRS End-to-end Access Time The mean value of GPRS End-to-end Access time measured by TEMS™ Investigation GSM was in this case 6,13 seconds. Measured mean values of appropriate performance indicators were:

• GPRS Attach Setup Time = 2,08 s. • PDP Context Activation time = 0,45 s. • GPRS Service Access Time = 3,61 s.

As it has already been mentioned in 4.3.3, the mean value of GPRS End-to-end Access Time, measured by TEMS™ Investigation GSM for every complete session access procedure is not the same value, as the sum of mean values of the GPRS Access chain’s elements (GPRS Attach PDP Context Activation GPRS Service Access), calculated by Equation 2.8.



5.1.3 GPRS Mean User Data Rate Mean values of Mean User Data Rate measured by TEMS™ Investigation GSM were:

• Measured by Method A: 20,60 kbps; • Measured by Method B: 17,41 kbps

5.1.4 GPRS Data Transfer Cut-off Ratio During this test there were no broken data transfers. Therefore, Data Transfer Cut-off Ratio is equal to 0%. The same situation is about PDP Context Cut-off Ratio parameter.

5.1.5 Round trip time and packet loss PING session results are shown in Table 5.1. Round trip time was measured by TEMS™ Investigation GSM and packet loss ratio was calculated by Equation 4.1.

Table 5.1 Measured RTT and Packet loss, HTTP, dynamic PDCH configuration, normal cell traffic load

Packet size [bytes] Parameter

32 512 1024 Round trip time [ms] 627 1117 1699 Packet loss [%] 1 0 1

5.2 KPI measurements: HTTP, high cell traffic load, dynamic PDCH configuration The test was performed with cell capacity reduced to 6 traffic channels (see 4.3) and approximate cell traffic load of 3,34 erlangs.

5.2.1 GPRS Accessibility Failure Ratio In this case, 36 of initial 50 attempts contained Attach Request (this procedure is given in 1.3.1). These GPRS Unavailability cases were caused by high traffic in the cell. Logfile further analysis has shown, that due to high traffic the MS could not get a free traffic channel for a certain amount of time. Finally, the Channel Request Timeout (T3126, see 3.1.2.3) had been reached. With totally 14 failed attempts to access GPRS service, the GPRS Unavailability in this case is equal to 28% (see Equation 2.3). This shows, what impact can high traffic have on GPRS availability in a cell. Besides, there was an unsuccessful attempt to Access GPRS Service. The reason is same as the reason in case described in 5.1.1 and, therefore, not related to radio network factors. Accessibility Failure Ratio, in this case, can be calculated as (see Equation 2.1):



[ ] ( ) %30100)0278,01()01()01()28,01(1% ≈×−×−×−×−−=HTTPRatioityFailureAccessibil





5.2.4 GPRS Data Transfer Cut-off Ratio 33 of 35 sessions were successfully completed (without data transfer cut-offs), while 2 sessions were broken. In both cases the traffic channel had been overtaken for voice communication. For a certain amount of time there was no traffic channel available and channel request timeout had been finally expired. In both cases, as well, the PDP context deactivation was not initiated by the MS. Therefore, GPRS Data Transfer Cut-off Ratio (see Equation 2.10) and PDP Context Cut-off Ratio (see Equation 2.11) are both equal to 5,714%.

5.2.5 Round trip time and packet loss PING session results are shown in Table 5.2. Round trip time was measured by TEMS™ Investigation GSM and packet loss ratio was calculated by Equation 4.1:

Table 5.2 Measured RTT and Packet loss, HTTP, dynamic PDCH configuration, high cell traffic load



5.3 KPI measurements: HTTP, high cell traffic load, combined PDCH configuration The test was performed with the same cell capacity as in 5.2 and one traffic channel dedicated for packet data, shared between two GPRS users- one using HTTP and another using WAP (see 4.3). Approximate traffic load was 3,4 erlangs.



5.3.1 GPRS Accessibility Failure Ratio One of 50 GPRS service attempts was not finished successfully. The reason of this failure is same as the reason in cases described in 5.1.1 and 5.2.1, and, therefore, it is not related to radio network factors. Accessibility Failure Ratio, in this case, can be calculated as (see Equation 2.1):

[ ] ( ) %2100)02,01()01()01()01(1% =×−×−×−×−−=HTTPRatioityFailureAccessibil





5.3.4 GPRS Data Transfer Cut-off Ratio During this test there were no broken data transfers. Therefore, Data Transfer Cut-off Ratio is equal to 0%. The same situation is about PDP Context Cut-off Ratio parameter.

5.3.5 Round Trip Time and Packet Loss PING session results are shown in Table 5.3. Round trip time was measured by TEMS™ Investigation GSM and packet loss ratio was calculated by Equation 4.1:

Table 5.3 Measured RTT and Packet loss, HTTP, combined PDCH configuration, high cell traffic load

Packet size [bytes] Parameter 32 548 1472

Round trip time [ms] 605 1263 1869 Packet loss [%] 0 0 2



5.4 KPI measurements: WAP, normal cell traffic load, dynamic PDCH configuration The test was performed in normal load conditions (see 4.3), with approximate cell traffic load value of 2,691 erlangs.

5.4.1 GPRS Accessibility Failure Ratio All of the 50 attempts to establish a WAP session over GPRS were finished successfully. Therefore, GPRS Accessibility Failure Ratio for WAP, in this case, is equal to 0%.

5.4.2 GPRS End-to-end Access Time The mean value of GPRS End-to-end Access time measured by TEMS™ Investigation GSM was in this case 4,88 seconds. Measured mean values of appropriate performance indicators were: GPRS Attach Setup Time = 2,14 s. PDP Context Activation time = 0,71 s. GPRS Service Access Time = 0,64 s. GPRS WAP Activation Time = 1,83 s.



5.4.4 GPRS Data Transfer Cut-off Ratio There were 4 data transfer cut-offs. In all 4 cases these cut-offs were caused not by radio network, but by WAP server. In every case the PDP Context Deactivation was initiated by the MS, after WAP session had been terminated. Therefore, Data Transfer Cut-off Ratio is equal to 8% (see Equation 2.10) and PDP Context Cut-off ratio is equal to 0%.


Table 5.4 Measured RTT and Packet loss, WAP, dynamic PDCH configuration, normal cell traffic load





5.5 KPI measurements: WAP, high cell traffic load, dynamic PDCH configuration The test was performed with cell capacity reduced to 6 traffic channels (see 4.3) and approximate cell traffic load of 3,7 erlangs.

5.5.1 GPRS Accessibility Failure Ratio Here the results are quite similar to the results in 5.2.1. 35 (of initial 50) attempts to access GPRS were finished successfully, with the same reason, as in 5.2.1 (channel request timeout). There was also one failed PDP Context Activation attempt. Logfile analysis has shown, that after successful Attach procedure, the traffic channel had been overtaken by a voice user. The MS did not manage to get a new traffic channel before channel assignment timeout had been finally reached. Therefore, in this case GPRS Unavailability is equal to 30% (see Equation 2.3) and PDP Context Activation Failure Ratio is equal to 2,857% (see Equation 2.5). GPRS Accessibility Failure Ratio can be calculated (according to Equation 2.2) as:

[ ] ( ) %32100)01()01()02857,01()01()3,01(1% ≈×−×−×−×−×−−=WAPRatioityFailureAccessibil


• GPRS Attach Setup Time = 2,92 s. • PDP Context Activation time = 0,66 s. • GPRS Service Access Time = 0,68 s. • GPRS WAP Activation Time = 2,10 s.



5.5.4 GPRS Data Transfer Cut-off Ratio There were 5 data transfer cut-offs, 4 of them were caused by PDP Context cut-off. The case, in which there was no PDP Context Cut-off was similar to those mentioned in 5.4.4 (the session was terminated from WAP server). The remaining 4 data transfer cut-offs were caused by an overtaken traffic channel and channel request timeout, as it was in case of failed PDP Context Activation mentioned in 5.5.1.



Therefore, according to Equations 2.10 and 2.11, Data Transfer Cut-Off Ratio and PDP Context Cut-Off Ratio are correspondingly equal to 14,706% and 11,765%.

5.5.5. Round Trip Time and Packet Loss PING session results are shown in Table 5.5. Round trip time was measured by TEMS™ Investigation GSM and packet loss ratio was calculated by Equation 4.1:

Table 5.5 Measured RTT and Packet loss, WAP, dynamic PDCH configuration, high cell traffic load



5.6 KPI measurements: WAP, high cell traffic load, combined PDCH configuration The test was performed with the same cell capacity as in 5.5 and one traffic channel dedicated for packet data, shared between two GPRS both using WAP (see 4.3). Approximate traffic load was 3,5 erlangs.

5.6.1 GPRS Accessibility Failure Ratio All of the 50 attempts to establish a WAP session over GPRS were finished successfully. Therefore, GPRS Accessibility Failure Ratio for WAP, in this case, is equal to 0%. It is similar to the results in 5.4.1.


• GPRS Attach Setup Time = 3,03 s. • PDP Context Activation time = 0,68 s. • GPRS Service Access Time = 0,86 s. • GPRS WAP Activation Time = 1,86 s.

5.6.3 Mean User Data Rate Mean values of Mean User Data Rate measured by TEMS™ Investigation GSM were:




5.6.4 Data Transfer Cut-Off Ratio There were 3 data transfer cut-offs. In all 3 cases these cut-offs were caused not by radio network, but by WAP server, as it was in 5.4.4. In every case the PDP Context Deactivation was initiated by the MS, after WAP session had been terminated. Therefore, Data Transfer Cut-off Ratio is equal to 6% (see Equation 2.10) and PDP Context Cut-off ratio is equal to 0%.


Table 5.6 Measured RTT and Packet loss, WAP, combined PDCH configuration, high cell traffic load





5.7 Comparison of tests’ results After the tests have been performed, we can compare the values of the key performance indicators. The results are given separately in tables and charts for HTTP (see 5.7.1) and for WAP (see 5.7.2).

5.7.1 HTTP Comparison of key performance indicators measured for HTTP is given in Table 5.7 and in charts (see Figures 5.1-5.5).

Table 5.7 Comparison of key performance indicators, measured for HTTP

PDCH configuration versus cell traffic load

Parameters Dynamic, normal Dynamic, high Combined, high Accessibility Failure Ratio [%] 4 30 2 Unavailability [%] 0 28 0 Attach Failure Ratio [%] 0 0 0 PDP Context Failure Ratio [%] 2 0 0 Service Access Failure Ratio [%] 2,04 2,78 2 End-To-End Access Time [s] 6,13 6,3 6,12 Attach Setup Time [s] 2,08 2,83 2,05 PDP Context Activation Time [s] 0,45 0,47 0,47 Service Access Time [s] 3,61 3,76 3,72 Mean User Data Rate (A) [kbit/s] 20,6 13,89 14,79 Mean User Data Rate (B) [kbit/s] 17,41 12,05 12,33 Data Transfer Cut-Off Ratio [%] 0 5,71 0 PDP Context Cut-Off Ratio [%] 0 0 0

5.7.1.1 Accessibility Failure Ratio The tests have shown significant increase of Accessibility Failure Ratio in case of dynamic PDCH configuration and high cell traffic load (see Figure 5.1, a), to be compared with two other scenarios.



Figure 5.1 Accessibility Failure Ratio for HTTP (charts): general values (A) and performance indicators’ values (B)

The performance indicator, that mostly forms this increase, is GPRS Unavailability caused by continious high traffic, and, as a result, expiry of Channel Request Timeout for MS in 14 cases (see 5.2.1). PDP Context Failire Ratio case (see 5.1.1) shows how interference can affect performance of GPRS service, even if there was no special interference tests intended and there were good interference conditions (sufficiently high C/I, ~ 20 dB) in the test environment (see 4.3). In fact, there was just one case, where interference level was so high, and GPRS service access had not been finished successfully due to interference conditions. In every test scenario, as well, there was a GPRS Service Access Failure (see 5.1.1, 5.2.1 and 5.3.1), that was, however, not related to radio network factors. It can be used for further researches of other factors’ impact on GPRS key performance indicators.

5.7.1.2 End-to-end Access Time Differently from Accessibility Failure Ratio, tests have shown, that End-to-end Access Time is not significantly affected by changes in PDCH configuration and cell traffic load. This can be seen from the Figure 5.2, A. If we take a look into the elements of GPRS Access chain (GPRS Attach

PDP Context Activation GPRS Service Access), we can see, that GPRS Service Access has the longest duration among the elements of this chain. It can be explained by the fact that at this stage there is communication with a server, which does not belong to EMT network. This inter-network communication therefore takes more time than all the other elements of GPRS Access chain for HTTP.



Figure 5.2 GPRS End-to-end Access Time for HTTP (charts): general values (A) and performance indicators’ values (B)

5.7.1.3 Mean User Data Rate The tests have shown, that Mean Data Rate is noticeably affected by cell traffic load (see Figure 5.3). We can also see, that in high load conditions we can acheive the same Mean User Data Rate with usage of a dedicated PDCH by 2 packet data users (one using HTTP and another using WAP) as for 1 packet data user without any dedicated PDCH.

05

10152025

Dynamic,normal

Dynamic, high Combined,high

PDCH configuration vs cell load

[kb

it/s

]

Mean User Data Rate (A) [kbit/s]

Mean User Data Rate (B) [kbit/s]

Figure 5.3 Mean User Data Rate for HTTP (chart)

5.7.1.4 Data Transfer Cut-off Ratio The tests have shown significant increase of Data Transfer Cut-off Ratio in case of dynamic PDCH configuration and high traffic load (see Figure 5.4), to be compared with two other scenarios.



Data Transfer Cut-Off Ratio [%]

02468

10

Dynamic,normal

Dynamic, high Combined,high


[%]


Figure 5.4 Data Transfer Cut-off Ratio for HTTP (chart)

5.7.1.5 Round Trip Time and Packet Loss The tests have shown, that round trip time depends on the packet length: the longer packet is, the more is RTT (see Figure 5.5, A). In case of dynamic PDCH configuration and high traffic load RTT is the longest. This is caused by two factors. First, with high traffic load in the cell there was just one traffic channel available. Second, overtaken traffic channel had led to ping timeouts and, as a result, increase in delay after another traffic channel became available for GPRS. In case of combined PDCH configuration and high cell traffic load RTT is smaller than in the previous case and it is caused by the absence of the factor of overtaken channels. It is, however, bigger than in the first case (dynamic PDCH configuration versus normal cell traffic load), because there was still just one traffic channel available for packet data communication with high traffic load in the cell, while in the first case there were more traffic channels available. Measurements have also shown, that in case of dynamic PDCH configuration and high traffic load in a cell there is significant increase of Packet Loss Ratio (see Figure 5.5, B). Lost packets have also been caused by overtaken traffic channels and ping timeouts (see previous paragraph).

Figure 5.5 RTT (A) and Packet Loss Ratio (B) for HTTP (charts)



5.7.2 WAP Comparison of key performance indicators measured for WAP is given in Table 5.8 and in charts (see Figures 5.6-5.10).

Table 5.8 Comparison of key performance indicators, measured for WAP

PDCH configuration versus cell traffic load

Parameters Dynamic, normal Dynamic, high Combined, high Accessibility Failure Ratio [%] 0 32 0 Unavailability [%] 0 30 0 Attach Failure Ratio [%] 0 0 0 PDP Context Failure Ratio [%] 0 2,857 0 Service Access Failure Ratio [%] 0 0 0 WAP Service Acess Failure Ratio [%] 0 0 0 End-To-End Access Time [s] 4,88 5,68 5,75 Attach Setup Time [s] 2,14 2,79 3,03 PDP Context Activation Time [s] 0,71 0,66 0,68 Service Access Time [s] 0,64 0,67 0,86 WAP Service Acess Time [s] 1,83 2,1 1,86 Mean User Data Rate (A) [kbit/s] 4,65 4,34 4,2 Mean User Data Rate (B) [kbit/s] 4,56 4,25 4,11 Data Transfer Cut-Off Ratio [%] 8 14,70588235 6 PDP Context Cut-Off Ratio [%] 0 11,76470588 0

5.7.2.1 Accessibility Failure Ratio The tests have shown significant increase of Accessibility Failure Ratio in case of dynamic PDCH configuration and high traffic load (see Figure 5.6, a), to be compared with two other scenarios. As in case with HTTP, the performance indicator, that mostly forms this increase, is also GPRS Unavailability, which has also same cause as HTTP- channel request timeout due to high load in a cell (see 5.4.1). An overtaken traffic channel has also caused a PDP Context Activation Failure (see 5.4.1).



Figure 5.6 Accessibility Failure Ratio for WAP (charts): general values (A) and performance indicators’ values (B)

We can also notice, that in cases of normal load/dynamic configuration and high load/combined configuration all the attempts to create a WAP session were finished successfully (Accessibility Failure Ratio = 0% in both cases). We cannot say the same about HTTP tests. However, as it was stated before (see 5.7.1.1), these failures were not caused by radio network factors.

5.7.2.2 End-to-end Access Time The tests have shown, that, like in case of HTTP, changes in PDCH configuration and cell traffic load have no significant impact on End-to-end Access Time for WAP (see Figure 5.7, A). One noticeable thing is a bit longer Attach procedure in case of high load (see Figure 5.8, B).

Figure 5.7 GPRS End-to-end Access Time for WAP (charts): general values (A) and performance indicators’ values (B)



5.7.2.3 Mean User Data Rate The tests have shown, that there was practically no impact on Mean User Data Rate (see Figure 5.8), unlike in case of HTTP. The difference here is in a smaller data rate used by WAP over GPRS.

00,5

11,5

22,5

33,5

44,5

5

Dynamic, normal Dynamic, high Combined, high


[kb

it/s

]

Mean User Data Rate (A) [kbit/s] Mean User Data Rate (B) [kbit/s]

Figure 5.8 Mean User Data Rate for WAP (chart)

5.7.2.4 Data Transfer Cut-off Ratio The tests have shown significant increase of Data Transfer Cut-off Ratio in case of dynamic PDCH configuration and high traffic load (see Figure 5.9), to be compared with two other scenarios.


0

2

4

6

810

12

14

16

18

20

Dynamic, normal Dynamic, high Combined, high


[%]


Figure 5.9 Data Transfer Cut-off Ratio for WAP (chart) However, unlike it was in case of HTTP tests, zero Data Transfer Cut-off Ratio could not be acheived. From the other hand, these cut-offs were caused not by radio network factors, but by the behaviour of WAP server (see 5.4.4). Therefore, even if we take into count only radio network-related cut-offs, then the impact on Data Transfer Cut-off Ratio is still significant.



5.7.2.5 Round Trip Time and Packet Loss Ratio The situation about Round Trip Time is the same, as with HTTP (see 5.7.1.5). You can see the results of the tests in Figure 5.10, A. However, we can see, that in WAP case there is bigger Packet Loss Ratio in normal cell traffic load/dynamic PDCH configuration and high cell traffic load/combined PDCH configuration conditions (see Figure 5.10, B) than in the same conditions versus HTTP case (for comparison, see Figure 5.5, B).

Figure 5.10 RTT (A) and Packet Loss Ratio (B) for WAP (charts)



CONCLUSIONS In this work I have analysed the impact of two radio network factors on the quality of service of GPRS applications (HTTP and WAP) in a particular GSM network. These two factors are bandwidth availability and co-channel interference. Bandwidth availability is defined by PDCH configuration and traffic load in a particular cell. The tests have shown that dynamic PDCH configuration can give good GPRS performance only in case of normal traffic load in a cell, when there are enough free traffic channels. But in case of a higher voice traffic load this does not work well due to the fact that traffic channels are being overtaken upon request from those subscribers who want to establish a circuit-switched connection for speech communication, which has a higher priority in GSM networks. This seriously affects three main QoS aspects for GPRS applications: Service Accessibility, Service Retainability and Service Integrity. Service Accessibility was affected for both WAP and HTTP services almost equally. In case of intensive voice traffic these services were much harder to be accessed with dynamic PDCH configuration in a cell than with combined channel configuration (one dedicated PDCH). The main indicator was GPRS Unavailability in the cell. Tests with WAP have shown excellent accessibility both in cases of normal cell traffic load/dynamic PDCH configuration and high cell traffic load/combined PDCH configuration. For WAP, there were no accessibility failure cases caused by non-radio network factors, while HTTP accessibility was also affected by non-radio network factors. Service Reiatnability was also affected for WAP and HTTP services. In case of intensive voice traffic there were more broken data trafsners with dynamic PDCH configuration in a cell than with combined channel configuration (one dedicated PDCH). In fact, there were no broken data transfers caused by radio network factors in both cases of normal cell traffic load/dynamic PDCH configuration and high cell traffic load/combined PDCH configuration. Tests with HTTP have shown excellent retainability in these cases, while WAP retainability was also affected by non-radio network factors. These factors are related to a particular WAP server. Service Integrity was also affected. The tests have shown that increase of traffic in the cell resulted into significant decrease of HTTP throughput in both dynamic and combined PDCH configuration cases. This can be explained by the fact that just one channel was available in both cases. WAP throughput has not been significantly affected by traffic increase. Tests with combined configuration and high traffic load have shown, that one HTTP and one WAP user can effectively share one dedicated PDCH. Throughput, in this case, will be the same as in case of dynamic configuration and high traffic in a cell. PING tests have shown Round Trip Time dependence on the packet length: longer packets cause longer delay. In case of dynamic PDCH configuration and high cell traffic load there was the longest RTT measured both for HTTP and WAP services, as well as the biggest value of Packet Loss Ratio. This shows that with combined PDCH configuration we can acheive much more reliable transmission of packet data even in case of longer delay caused by intensive voice traffic in a cell. Generally, modifying the PDCH configuration of a particular cell can reduce the impact of bandwidth availability factor on the QoS of GPRS applications. Combined configuration is effective in intensive voice traffic conditions, because in this case we acheive:

• The same accessibility and reliability for both HTTP and WAP, if we compare it to dynamic PDCH configuration in normal voice traffic conditions,

• The same throughput for each service, with one dedicated PDCH shared by two users, if we compare it to dynamic PDCH configuration in intensive voice traffic conditions with one GPRS user,



• More reliable transmission of packet data even in case of longer delay caused by intensive voice traffic in a cell. To say more, in conditions of high cell traffic load, there is still smaller delay with combined PDCH configuration than with dynamic PDCH configuration.

To modify the PDCH configuration of a particular cell it is necessary to collect the cell’s traffic load statistics data. Special attention must be paid to those cells that are located in central (business) town districts, where GPRS service is more frequently used and mean time voice traffic is sufficiently high to be an obstacle for packet data communication. Differing from bandwidth availability factor, the impact of interference on GPRS quality of service could not be completely defined. This is due to the fact that there was no possibility to test link adaptation procedures with GPRS coding schemes, because this feature was not available in the tested network. We should also consider the fact that the tests were performed in a cell with good interference conditions. That is why there was just one fault case caused by high level of interference. And it also shows, that without link adaptation it is normal that retransmission with the same coding scheme can not be performed when current C/I is below the threshold level specified for this scheme. This work covers two radio network factors that have their impact on GPRS QoS. But there are more factors that also affect quality of packet data services in GSM networks and their impact also needs to be defined. The subjects for future studies can be:

• Impact of radio network factors on EDGE performance. One special moment in this case is to define the impact of interference on EDGE performance, considering link adaptation with EDGE channel coding schemes;

• Impact of user mobility on GPRS/EDGE performance. • Impact of non-radio network factors on GPRS/EDGE performance. These factors can be,

for example, PCU capacity and SGSN <-> GGSN communication.



REFERENCES [1] - Ericsson AXE Documentation Library, Serving GPRS Support Node, GPRS Overview [2] - Siemens mobile, “EDGE: Supercharging your GSM network for high-speed mobile data”, http://communications.siemens.com/repository/730/73089/EDGE_0703.pdf [3] – Ericsson White Paper Technical, “EDGE: Introduction of high-speed data in GSM/GPRS networks” http://www.ericsson.com/products/white_papers_pdf/edge_wp_technical.pdf [4] –Karam Jinane, Peignot Estelle, Abou Aun Mayssam: GPRS/EDGE presentation, http://dessr2m.adm-eu.uvsq.fr/gprs-edge.pdf [5] - Lars Ekeroth, Per-Martin Hedström, “GPRS support nodes”, pages 157-160, http://www.ericsson.com/about/publications/review/2000_03/files/2000034.pdf [6] - ITU-T recommendation E.800 (1994), Terms and definitions related to quality of service and network performance including dependability [7] - The IEEE paper “Distributed Multimedia and Quality of Service” [8] - 3GPP2 S.R0035, Quality of Service, Stage 1 – Requirements [9] - Nicolae D. Cotanis, LCC International, Inc., R&D Dept.McLean, USA. “QoS Estimation for cellular packet data networks”. http://www.lcc.com/media_center/QoS%20Estimation%20for%20Cellular%20Packet%20Data%20Networks.doc.pdf [10] - P. Francis-Cobley, N. Davies, Department of Computer Science, University of Bristol “Quality of Service Issues in Heterogeneous Network Systems”, slide 29, http://scitec.uwichill.edu.bb/cmp/online/cs31k/Quality%20of%20Service.ppt [11] – Global Knowledge TM “Expert Series”, “Introduction to Quality of Service”. Pages 2-3. http://whitepapers.zdnet.co.uk/0,39025945,60068376p-39000619q,00.htm [12] - ETSI TS 102 250-2 V1.2.1 (2004-06), Speech Processing, Transmission and Quality Aspects (STQ); QoS aspects for popular services in GSM and 3G networks; Part 2: Definition of Quality of Service parameters and their computation. [13] – Ericsson TEMS™ Investigation GSM User Manual: GPRS Key Performance Indicators [14] - Ivar Jørstad, ” General Packet Radio Service (GPRS)”, Mobile Telematics 2004 http://www.item.ntnu.no/fag/tm8100/Pensumstoff2004/GPRSIVAR.ppt [15] – Source O2, GPRS: Reliability, Latency and Jitter. http://www.sourceo2.com/O2_Developers/O2_technologies/GPRS/Technical_overview/gprs_tech_reliability_latency_jitter.htm [16] - Focus on GPRS http://www.cellular.co.za/gprs.htm [17] – Ericsson Hrvatska, GPRS http://www.ericsson.hr/etk/revija/br_2_2001_ru/gprs_ru.htm [18] - Dynamic Telecommunications, Inc. “Co-Channel Interference in Wireless Networks. Effects and Methods of Measurement”, May 2000. Page2. http://www.dynatele.com/WhitePapers/Co-ChannelInterference.pdf


http://communications.siemens.com/repository/730/73089/EDGE_0703.pdf

http://www.ericsson.com/products/white_papers_pdf/edge_wp_technical.pdf

http://dessr2m.adm-eu.uvsq.fr/gprs-edge.pdf

http://www.ericsson.com/about/publications/review/2000_03/files/2000034.pdf

http://www.lcc.com/media_center/QoS Estimation for Cellular Packet Data Networks.doc.pdf

http://www.lcc.com/media_center/QoS Estimation for Cellular Packet Data Networks.doc.pdf

http://scitec.uwichill.edu.bb/cmp/online/cs31k/Quality of Service.ppt

http://whitepapers.zdnet.co.uk/0,39025945,60068376p-39000619q,00.htm

http://www.item.ntnu.no/fag/tm8100/Pensumstoff2004/GPRSIVAR.ppt

http://www.sourceo2.com/O2_Developers/O2_technologies/GPRS/Technical_overview/gprs_tech_reliability_latency_jitter.htm

http://www.sourceo2.com/O2_Developers/O2_technologies/GPRS/Technical_overview/gprs_tech_reliability_latency_jitter.htm

http://www.cellular.co.za/gprs.htm

http://www.ericsson.hr/etk/revija/br_2_2001_ru/gprs_ru.htm

http://www.dynatele.com/WhitePapers/Co-ChannelInterference.pdf

http://www.dynatele.com/WhitePapers/Co-ChannelInterference.pdf


[19] - Juan Li: Link Adaptation in General Packet Radio Services (GPRS); Helsinki University of Technology, 28th.May, 2002 http://www.comlab.hut.fi/opetus/158/WebMateriaali/li_280502.pdf [20] – Alexander Brizmer: “MMS teenusekvaliteeti mõjutavad faktorid GSM raadiovõrgus”, Tallinn Technical University, Chair of Signal Processing, 2003. [21] - Queseth, O.; Gessler, F.; Frodigh, M. “Algorithms for link adaptation in GPRS.”, Vehicular Technology Conference, 1999 IEEE 49th, Volume: 2 , 1999 Page(s): 943 -947 vol.2. [22] - William Stallings, “Data and Computer Communications”, fourth edition, Macmillan publishing company, 1994. [23] – Cerritos Linux User Group, Tutorials: Hugo Samayoa, “GRE Tunneling”, http://www.cerritoslug.org/tutorials/gre_tunneling.html [24] - Tunneling for Dollars: Comparing IPSec and PPTP for Extranet Security By Julie Bort http://www.intranetjournal.com/foundation/tunneling.shtml [25] – Ericsson TEMS™ Investigation GSM white paper, http://www.ericsson.com/services/tems/downloads/ds_investigation_gsm_5.1.pdf [26] – Agilent Technologies, Agilent OSS acceSS7. http://we.home.agilent.com/USeng/nav/-536885432.0/pc.html [27] - Ericsson AXE Documentation Library, CPI for BSS R10 MD, Circuit-switched Traffic Timers [28] – Electronics Research Group: Selective Repeat Error Recovery, http://www.erg.abdn.ac.uk/users/gorry/course/arq-pages/sr.html [29] – Ericsson EDGE White paper Technical, http://www.ericsson.com/products/white_papers_pdf/edge_wp_technical.pdf [30] – Amobile Классификация и типы телефонов с GPRS, http://www.amobile.ru/gprs/class.htm [31] – Andrei Gurtov, Reiner Ludwig, Evaluating the Eifel Algorithm for TCP in a GPRS network, http://www.cs.helsinki.fi/u/gurtov/papers/ew02.pdf


http://www.comlab.hut.fi/opetus/158/WebMateriaali/li_280502.pdf

http://www.cerritoslug.org/tutorials/gre_tunneling.html

http://www.intranetjournal.com/foundation/tunneling.shtml

http://www.ericsson.com/services/tems/downloads/ds_investigation_gsm_5.1.pdf

http://we.home.agilent.com/USeng/nav/-536885432.0/pc.html

http://we.home.agilent.com/USeng/nav/-536885432.0/pc.html

http://www.erg.abdn.ac.uk/users/gorry/course/arq-pages/sr.html

http://www.ericsson.com/products/white_papers_pdf/edge_wp_technical.pdf

http://www.amobile.ru/gprs/class.htm


APPENDIX A: TWO METHODS OF MEAN DATA RATE MEASUREMENTS Currently two main views about the best way to reflect the user's experience are in place: One preferring the payload throughput philosophy and the other preferring the transaction throughput philosophy:

• Method A defines trigger points, which are as independent as possible from the service used, therefore representing a more generic view (payload throughput).

• Method B defines trigger points on application layer, therefore representing a more service-oriented view (transaction throughput).

An example of the different trigger points defined for each set is illustrated in Figure A.1 and Figure A.2. The application tested in this case is HTTP. The start trigger point for the Mean Data Transfer for Web browsing is either the reception of the first packet containing data content (Method A, see Figure A.1) or the sending of the HTTP GET command (Method B. See Figure A.2).

Figure A.1 Measurement of Mean Data Rate for HTTP, Method A [12]



Figure A.2 Measurement of Mean Data Rate for HTTP, Method B [12]



APPENDIX B: BLOCK STRUCTURES AND PARAMETERS OF GPRS CHANNEL CODING SCHEMES Figure B.1 shows the data segmentation flow between different layers.

Figure B.1 Data flow segmentation between layers [19] The resulting RLC data blocks are then coded (see Figure B.2) and block-interleaved over four normal bursts in consecutive TDMA frames.

Figure B.2 Encoding of GPRS data packets [14]



Figure B.3 shows radio block structures for GPRS channel coding schemes CS1-CS4.

Figure B.3 Radio block structures for GPRS coding schemes: a) CS1-CS3 b) CS4 [19] CS1 to CS3 are based on a half rate convolutional encoder. However, they differ on the puncturing schemes applied to the output of this encoder. Block Check Sequences are used in all the schemes to facilitate the error detection at the receiver. The characteristics of the different coding schemes are summarized in Table B.1.

Table B.1 Parameters of GPRS channel coding schemes [30]

Channel Coding Scheme CS-1 CS-2 CS-3 CS-4 Pre-cod. USF 3 6 6 12 Infobits without USF 181 268 312 428 Parity bits BC 40 16 16 16 Tail bits 4 4 4 - Output conv encoder 456 588 676 456 Punctured bits 0 132 220 - Code rate 1/2 ~2/3 ~3/4 1 Data rate kbit/s 9.05 13.4 15.6 21.4 Maximum data speed with 8 time-slots 72.4 kb/s 107.2 kb/s 124.8 kb/s 171.2 kb/s



APPENDIX C: LINK ADAPTATION ALGORITHMS In a real system, the C/I for a transmission that is being started is unknown and therefore has to be estimated. The issue is to identify an algorithm that uses some measurable parameter to approximate C/I and to choose the best coding scheme. To reach this goal, three questions have to be answered:

• What parameters should be used to estimate the carrier to interference ratio? • How should the algorithms determining the choice of coding scheme be designed? • What performance measures should be used to evaluate different algorithms?

The algorithms proposed here take into account three aspects of the choice of coding schemes:

• what coding scheme should be used initially, • how to update the chosen coding scheme, • what system parameter to measure.

C.1 Estimated C/I algorithm: The implementation of the CIR-based algorithm can be described as:

1. Starting with coding scheme CS-4. 2. The RLC blocks are continuously transmitted until the end of one reporting window (10

blocks was selected in the simulation). 3. Then the CIR is estimated by subspace-based estimator [4]. If the result exceeds

specified threshold (see Figure 3.7 for threshold values), the coding scheme is modified in the following data blocks and keeps unchanged during the following averaging window.

4. If there are more data to transmit go back to step 2. Failed blocks will be placed at the head of the queue, and are retransmitted first in the next reporting window.

C.2 Block error rate algorithm: The parameter used to estimate the channel quality is the Block Error Rate (BLER). The BLER at intersections are taken as the thresholds (see Figure 3.7) and they are calculated as:

−=

max

1ThrThrBLER

Equation C.1 Where Thr and Thrmax are current and maximum throughput values [19]. If the block error rate of the previous reporting window lies above a certain threshold, a stronger coding scheme should be used. If it instead lies below some other threshold, a weaker one should be chosen. The implementation of BLER-based algorithm can be described as:



1. Starting with coding scheme CS-4. 2. The RLC blocks are continuously transmitted until the end of one reporting window (10

blocks was selected in the simulation). 3. Then the BLER is averaged over the previous reporting window. If the result locates

outside the range of the current coding scheme defined by specified threshold, the coding scheme is modified in the following data blocks and keeps unchanged during the following averaging window.

4. If there are more data to transmit go back to step 2. Failed blocks will be placed at the head of the queue, and are retransmitted first in the next reporting window.

An example of BLER-based algorithm [19] is shown in Figure C.1.

Figure C.1 an example of BLER-based algorithm [19]



APPENDIX D: RETRANSMISSION WITH SELECTIVE ARQ Selective ARQ (Automatic Repeat Request) is a procedure, which is implemented in some communications protocols to provide reliability. It is the most complex of a set of procedures which may provide error recovery, it is however the most efficient scheme. Selective repeat is employed by the TCP transport protocol. The recovery of a corrupted Packet Data Unut (PDU) proceeds in four stages:

• First, the corrupted PDU is discarded at the remote node's receiver. • Second, the remote node requests retransmission of the missing PDU using a control

PDU (sometimes called a Selective Reject). The receiver then stores all out-of-sequence PDUs in the receive buffer until the requested PDU has been retransmitted.

• The sender receives the retransmission request and then transmits the lost PDU(s). • The receiver forwards the retransmitted PDU, and all subsequent in-sequence PDUs

which are held in the receive buffer.

Figure D.1 Retransmission using Selective Repeat [28]

A remote node may request retransmission of corrupted PDUs by initiating Selective Repeat error recovery by sending a control PDU indicating the missing PDU. This allows the remote node to instruct the sending node where to retransmit the PDU, which has not been received. The remote stores any out-of-sequence PDUs (i.e. which do not have the expected sequence number) until the retransmission is complete. Upon receipt of a Selective Repeat control PDU (by the local node), the transmitter sends a single PDU from its buffer of unacknowledged PDUs. The transmitter then continues normal transmission of new PDUs until the PDUs are acknowledged or another selective repeat request is received. Below is an example of Selective Repeat Operation (see Figure D.2).



Figure D.2 An example of retransmission with selective ARQ [28] The sender transmits four PDUs (1-4). The first PDU (1) is corrupted and not received. The receiver detects this when it receives PDU (2), which it stores in the receive buffer and requests a selective repeat of PDU (1). The sender responds to the request by sending PDU (1), and then continues sending PDUs (5-7). The receiver stores all subsequent out-of-sequence PDUs (3-4), until it receives PDU (1) correctly. The received PDU (1) and all stored PDUs (2-4) are then forwarded, followed by (5-7) as each of these is received in turn.



APPENDIX E: OVERVIEW OF EDGE (EGPRS) EDGE (also referred as Enhanced GPRS, EGPRS) is the next step in the evolution of GSM and IS-136. The objective of the new technology is to increase data transmission rates and spectrum efficiency and to facilitate new applications and increased capacity for mobile use. GPRS allows data rates of 115 kbps and, theoretically, of up to 160 kbps on the physical layer. EGPRS is capable of offering data rates of 384 kbps and, theoretically, of up to 473.6 kbps. A new modulation technique and error-tolerant transmission methods, combined with improved link adaptation mechanisms, make these EGPRS rates possible. This is the key to increased spectrum efficiency and enhanced applications, such as wireless Internet access, e-mail and file transfers. Basically, EDGE only introduces a new modulation technique and new channel coding that can be used to transmit both packet-switched and circuit-switched voice and data services. EDGE is therefore an add-on to GPRS and cannot work alone. GPRS has a greater impact on the GSM system than EDGE has. By adding the new modulation and coding to GPRS and by making adjustments to the radio link protocols, EGPRS offers significantly higher throughput and capacity. With EDGE, the same time slot can support more users. This decreases the number of radio resources required to support the same traffic, thus freeing up capacity for more data or voice services. EDGE makes it easier for circuit-switched and packet-switched traffic to coexist while making more efficient use of the same radio resources. Thus in tightly planned networks with limited spectrum, EDGE may also be seen as a capacity booster for the data traffic. GPRS and EGPRS have different protocols and different behavior on the base station system side. However, on the core network side, GPRS and EGPRS share the same packet-handling protocols and, therefore, behave in the same way. Reuse of the existing GPRS core infrastructure (serving GRPS support node/gateway GPRS support node) emphasizes the fact that EGPRS is only an “add-on” to the base station system and is therefore much easier to introduce than GPRS (Figure E.1).



Figure E.1 EGPRS introduces changes to GPRS only on the base station system part of the network [29]

E.1 Technical differences between GPRS and EGPRS Table E.1 compares the basic technical data of GPRS and EDGE. Although GPRS and EDGE share the same symbol rate, the modulation bit rate differs. EDGE can transmit three times as many bits as GPRS during the same period of time. This is the main reason for the higher EDGE bit rates. The differences between the radio and user data rates are the result of whether or not the packet headers are taken into consideration. These different ways of calculating throughput often cause misunderstanding within the industry about actual throughput figures for GPRS and EGPRS.

Table E.1 Comparison of GPRS and EDGE technical data [29]

GPRS EDGE Modulation GMSK 8-PSK/GMSK Symbol rate 270 ksym/s 270 ksym/s Modulation bit rate 270 kb/s 810 kb/s Radio data rate per time slot 22,8 kb/s 69,2 kb/s User data rate per time slot 20 kb/s (CS4) 59,2 kb/s (MCS9) User data rate (8 time slots) 160 kb/s 473,6 kb/s

The data rate of 384 kbps is often used in relation to EDGE. The International Telecommunications Union (ITU) has defined 384 kbps as the data rate limit required for a service to fulfill the International Mobile Telecommunications-2000 (IMT-2000) standard in a pedestrian environment. This 384 kbps data rate corresponds to 48 kbps per time slot, assuming an eight-time slot terminal.



E.1.1 EDGE modulation technique The modulation type that is used in GSM is the Gaussian minimum shift keying (GMSK), which is a kind of phase modulation. This can be visualized in an I/Q diagram that shows the real (I) and imaginary (Q) components of the transmitted signal (Figure E.2). Transmitting a zero bit or one bit is then represented by changing the phase by increments of ±p. Every symbol that is transmitted represents one bit; that is, each shift in the phase represents one bit.

Figure E.2 I/Q diagram showing EDGE modulation benefits [29]

To achieve higher bit rates per time slot than those available in GSM/GPRS, the modulation method requires change. EDGE is specified to reuse the channel structure, channel width, channel coding and the existing mechanisms and functionality of GPRS. The modulation standard selected for EDGE, 8-phase shift keying (8PSK), fulfills all of those requirements. 8PSK modulation has the same qualities in terms of generating interference on adjacent channels as GMSK. This makes it possible to integrate EDGE channels into an existing frequency plan and to assign new EDGE channels in the same way as standard GSM channels. The 8PSK modulation method is a linear method in which three consecutive bits are mapped onto one symbol in the I/Q plane. The symbol rate, or the number of symbols sent within a certain period of time, remains the same as for GMSK, but each symbol now represents three bits instead of one. The total data rate is therefore increased by a factor of three. The distance between the different symbols is shorter using 8PSK modulation than when using GMSK. Shorter distances increase the risk for misinterpretation of the symbols because it is more difficult for the radio receiver to detect which symbol it has received. Under good radio conditions, this does not matter. Under poor radio conditions, however, it does. The “extra” bits will be used to add more error-correcting coding, and the correct information can be recovered. Only under very poor radio environments is GMSK more efficient. Therefore the EDGE coding schemes are a mixture of both GMSK and 8PSK.

E.1.2 Coding schemes For GPRS, four different coding schemes, designated CS1 through CS4, are defined. Each has different amounts of error-correcting coding that is optimized for different radio environments. For



EGPRS, nine modulation coding schemes, designated MCS1 through MCS9, are introduced. These fulfill the same task as the GPRS coding schemes. The lower four EGPRS coding schemes (MSC1 to MSC4) use GMSK, whereas the upper five (MSC5 to MSC9) use 8PSK modulation. Figure E.3 shows both GPRS and EGPRS coding schemes, along with their maximum throughputs.

Figure E.3 Coding schemes for GPRS and EGPRS (user data rate per time slot) [29] GPRS user throughput reaches saturation at a maximum of 20 kbps with CS4, whereas the EGPRS bit rate continues to increase as the radio quality increases, until throughput reaches saturation at 59.2 kbps. Both GPRS CS1 to CS4 and EGPRS MCS1 to MCS4 use GMSK modulation with slightly different throughput performances. This is due to differences in the header size (and payload size) of the EGPRS packets. This makes it possible to resegment EGPRS packets. A packet sent with a higher coding scheme (less error correction) that is not properly received, can be retransmitted with a lower coding scheme (more error correction) if the new radio environment requires it. This resegmenting (retransmitting with another coding scheme) requires changes in the payload sizes of the radio blocks, which is why EGPRS and GPRS do not have the same performance for the GMSK- modulated coding schemes. Resegmentation is not possible with GPRS.

E.1.3 Packet handling Another improvement that has been made to the EGPRS standard is the ability to retransmit a packet that has not been decoded properly with a more robust coding scheme. For GPRS, resegmentation is not possible. Once packets have been sent, they must be retransmitted using the original coding scheme even if the radio environment has changed. This has a significant impact on the throughput, as the algorithm decides the level of confidence with which the link adaptation (LA) must work. As a result, the link adaptation for GPRS requires careful selection of the coding scheme in order to avoid retransmissions as much as possible.



With EGPRS, resegmentation is possible. Packets sent with little error protection can be retransmitted with more error protection, if required by the new radio environment. The rapidly changing radio environment has a much smaller effect on the problem of choosing the wrong coding scheme for the next sequence of radio blocks because resegmentation is possible. Therefore, the EGPRS link-controlling algorithm can be very aggressive when selecting the modulation coding schemes.

E.1.4 Addressing window Before a sequence of coded radio link control packets or radio blocks can be transmitted over the Um (radio) interface, the transmitter must address the packets with an identification number. This information is then included in the header of every packet. The packets in GPRS are numbered from 1 to 128. After transmission of a sequence of packets (e.g., 10 packets), the transmitter asks the receiver to verify the correctness of the packets received in the form of an acknowledged/unacknowledged report. This report informs the transmitter which packet or packets were not successfully decoded and must be retransmitted. Since the number of packets is limited to 128 and the addressing window is 64, the packet sending process can run out of addresses after 64 packets. If an erroneously decoded packet must be retransmitted, it may have the same number as a new packet in the queue. If so, the protocol between the terminal and the network stalls, and all the packets belonging to the same low-layer capability frame must be retransmitted. In EGPRS, the addressing numbers have been increased to 2048 and the window has been increased to 1024 in order to minimize the risk for stalling. This, in turn, minimizes the risk for retransmitting low-layer capability frames and prevents decreased throughput.

E.1.5 Interleaving To increase the performance of the higher coding schemes in EGPRS (MCS7 to MCS9) even at low C/I, the interleaving procedure has been changed within the EGPRS standard. When frequency hopping is used, the radio environment is changing on a per-burst level. Because a radio block is interleaved and transmitted over four bursts for GPRS, each burst may experience a completely different interference environment. If just one of the four bursts is not properly received, the entire radio block will not be properly decoded and will have to be retransmitted. With EGPRS, the standard handles the higher coding scheme differently than GPRS to combat this problem. MCS7, MCS8 and MCS9 actually transmit two radio blocks over the four bursts, and the interleaving occurs over two bursts instead of four. The likelihood of receiving two consecutive error-free bursts is higher than receiving four consecutive error-free bursts. This means that the higher coding schemes for EDGE have a better robustness with regard to frequency hopping.

E.1.6 EGPRS link controlling function To achieve the highest possible throughput over the radio link, EGPRS uses a combination of two functionalities: link adaptation and incremental redundancy. Compared to a pure link adaptation solution, this combination of mechanisms significantly improves performance. Link adaptation uses the radio link quality, measured either by the mobile station in a downlink transfer or by the base station in an uplink transfer, to select the most appropriate modulation coding scheme for transmission of the next sequence of packets. For an uplink packet transfer, the network informs the mobile station which coding scheme to use for transmission of the next sequence of packets. The modulation coding scheme can be changed for each radio block (four bursts), but a change is usually initiated by new quality estimates. The practical adaptation rate is therefore decided by the measurement interval.



There are three families: A, B and C. Within each family, there is a relationship between the payload sizes, which makes resegmentation for retransmissions possible. Incremental redundancy initially uses a coding scheme, such as MCS9, with very little error protection and without consideration for the actual radio link quality. When information is received incorrectly, additional coding is transmitted and then soft combined in the receiver with the previously received information. Soft combining increases the probability of decoding the information. This procedure will be repeated until the information is successfully decoded. For the mobile stations, incremental redundancy support is mandatory in the standard.

E.2 Impact of EGPRS on existing GSM/GPRS networks Due to the minor differences between GPRS and EGPRS, the impact of EGPRS on the existing GSM/GPRS network is limited to the base station system. The base station is affected by the new transceiver unit capable of handling EDGE modulation as well as new software that enables the new protocol for packets over the radio interface in both the base station and base station controller. The core network does not require any adaptations. Due to this simple upgrade, a network capable of EDGE can be deployed with limited investments and within a short time frame.


Date post:	21-Jul-2016
Category:	Documents
Upload:	stephen-nguyen
View:	25 times
Download:	2 times

radio network gprs qos

Documents