Download Op­ti­miz­ing Voice in ATM/ IP Mo­bile I Net­works

 Op­ti­miz­ing Voice in ATM/ IP Mo­bile I Net­works

(TELE­COM EN­GI­NEER­ING I)

Juli­et Bates- Mc­GrawHill

 Con­tents

 Fore­word

 Pref­ace

 Ac­knowl­edg­ments

 In­tro­duc­tion

 Chap­ter 1 Mo­bile Voice Ac­cess to ATM/ IP Net­works 1

 1.1 Sec­ond- ​and Third- ​Gen­er­ation Mo­bile Net­works Stan­dards 2

 1.2 ATM and

 IP Trans­port in UMTS 3

 1.3 ATM in UMTS 4

 1.4 The UTRAN

 1.5 Speech Cod­ing in the UTRAN 5

 1.6 QoS in IP Net­works 6

 1.7 Me­dia Gate­ways (MG­Ws) 7

 1.8 Me­dia Gate­way Con­troller 9

 1.9 In­ter­work­ing Be­tween IP and ATM 11

Chap­ter 2 PBX Voice Ac­cess to ATM 13

 2.1 Dis­ad­van­tages of In­her­it­ed PBX So­lu­tions 13

 2.2 Ad­van­tages of Mi­grat­ing Voice to ATM 14

 2.3 Vir­tu­al Pri­vate Net­works (VP­Ns) 5

 2.3.1 En­ter­prise VP­Ns 1

 2.3.2 PVC- ​Based En­ter­prise VP­Ns 15

 2.3.3 SVC- ​Based En­ter­prise VP­Ns 6

 2.4 G.711 Cod­ing 17

 2.5 Sig­nal­ing In­for­ma­tion 18

 2.5.1 Chan­nel As­so­ci­at­ed Sig­nal­ing (CAS) 1

 2.5.2 QSIG and DP­NSS 9

 2.5.3 Com­mon Chan­nel Sig­nal­ing (CCS) 1

 2.5.4 Pri­vate Sig­nal­ing Sys­tem No.1 (PSS1) 19

 2.5.5 Nar­row­band IS­DN and Q.931/ DSS1

 2.6 Trans­port­ing PBX Sig­nal­ing 20

 2.6.1 Trans­port­ing Sig­nal­ing Trans­par­ent­ly 2

 2.6.2 In­ter­work­ing Sig­nal­ing 21

 Chap­ter 3 Math­emat­ical Mod­el­ing of Voice Traf­fic 23

 3.1 Sta­tis­ti­cal Gain in the ATM Core 23

 3.2 Mea­sure­ment of Con­ver­sa­tion­al Voice 24

 3.2.1 Gru­ber's Re­sults 25

 3.2.2 Brady's Re­sults 6

 3.2.3 Sri­ram's Re­sults

 3.2.4 Heffes and Lu­can­toni's Re­sults 27

 3.2.5 Sri­ram and Whitt's Re­sults 8

 3.2.6 Cheng's Re­sults 31 Con­clu­sions 33

 Chap­ter 4 Trans­port­ing Voice over ATM 35

 4.1 Adap­ta­tion to ATM 35

 4.2 Al­ter­na­tive Adap­ta­tion Lay­ers for Voice over ATM 35

 4.3 Cir­cuit Em­ula­tion Us­ing AAL1 36

 4.3.1 AAL1 Pack­et Da­ta Unit (PDU) For­mat 3

 4.3.2 Un­struc­tured Cir­cuit Em­ula­tion 37

 4.3.3 Struc­tured Cir­cuit Em­ula­tion 8

 4.4 Band­width Re­quire­ments for Cir­cuit Em­ula­tion 3

 4.4.1 Cir­cuit Em­ula­tion in Un­struc­tured Mode 8

 4.4.2 Cir­cuit Em­ula­tion in Struc­tured Mode 39

 4.5 Voice­band Ser­vices Us­ing AAL2 41

 4.5.1 AAL2 Com­mon Part Sub­lay­er (CPS) 43

 4.6 Voice­band Ser­vices Us­ing AAL5 45

 4.6.1 Voice and Da­ta With­in an FRF.11 Sub­frame 4 Con­clu­sions 47

 Chap­ter 5 Voice­band Pro­cess­ing 49

 5.1 Com­pres­sion 50

 5.2 Clear Chan­nel 51

 5.3 Com­pres­sion Coder Units 5

 5.4 Com­bin­ing Cod­ed Units in an AAL Frame 51

 5.4.1 Com­bin­ing Coder Units in­to an AAL2 Frame 53

 5.4.2 Com­bin­ing Cod­ed Units in­to an AAL5 Frame 56

 5.5 Com­par­ison of AAL1, AAL2, and AAL5 Band­width 58

 5.6 Com­par­ison of De­lay in AAL1, AAL2, and AAL5

 5.7 Di­verse Band­width Re­quire­ments of a Sin­gle Voice Source 63

 5.8 Si­lence De­tec­tion and Sup­pres­sion 64

 5.9 Ben­efits of Si­lence Sup­pres­sion

 5.10 Si­lence De­tec­tion Over­heads 65 Con­clu­sions 66

 Chap­ter 6 Qual­ity of Ser­vice in an ATM Net­work 69

 6.1 Traf­fic Man­age­ment in an ATM Net­work 69

 6.2 Ser­vice Cat­egories and QoS Class­es 70

 6.2.1 Con­stant Bit Rate (CBR) 1

 6.2.2 Vari­able Bit Rate (VBR) 72

 6.2.3 Avail­able Bit Rate (ABR)

 6.2.4 Un­spec­ified Bit Rate (UBR) 72

 6.3 ATM QoS Class­es 7

 6.4 ATM QoS Class Pa­ram­eters 3

 6.4.1 Cell Loss Ra­tios (CLRs) With­in a QoS Class 73

 6.4.2 Cell Trans­fer De­lay (CTD) With­in a QoS Class 74

 6.4.3 Cell De­lay Vari­ation (CDV) With­in a QoS Class 75

 6.4.4 End- ​to- ​End CTD and CDV 76

 6.5 Re­la­tion­ship Be­tween Load, CLR, and CDV in a Sin­gle Shared Buffer 77

 6.6 QoS Tar­gets 78

 6.7 Traf­fic- ​De­scrip­tor Pa­ram­eters 79

 6.8 Call Ad­mis­sion Con­trol (CAC) Al­go­rithm 80

 6.8.1 Steady State Queue Be­hav­ior

 6.8.2 Defin­ing the Ef­fec­tive Band­width 81

 6.9 Over­book­ing 8 Con­clu­sions 2

 Chap­ter 7 The Voice Mod­el 85

 7.1 Char­ac­ter­iza­tion of Voice 86

 7.2 Uni­form Ar­rival and Ser­vice Mod­el 87

 7.3 No­ta­tion of Queu­ing Sys­tems 8

 7.4 Queue Emp­ty­ing 89

 7.5 Stochas­tic Mod­el­ing of a Sin­gle Queue in an ATM Switch 89

 7.5.1 Mod­el­ing at the Talk­spurt Lev­el 90

 7.5.2 The Ef­fects of Shap­ing 91

 7.6 Siz­ing the Buffer 92

 7.6.1 Meet­ing a De­lay Tar­get 9 Con­clu­sions 3

 Chap­ter 8 Traf­fic Mod­el­ing at the Talk­spurt Lev­el 95

 8.1 Life­time of a Talk­spurt 95

 8.2 Marko­vian Meth­ods of Mod­el­ing the Buffer Queue 96

 8.2.1 Con­tin­uous Time Markov Chain 9

 8.2.2 Prob­abil­ity of the Last Cell

 8.2.3 Birth and Death Pro­cess 98

 8.2.4 A Pure Birth or Death Pro­cess 99

 8.3 Es­ti­mat­ing the Num­ber of Calls Con­cur­rent­ly in a Talk­spurt 99

 8.4 Ef­fec­tive Band­width 100

 8.5 A Method for Di­men­sion­ing the MBS 101

 8.6 Set­ting the PCR, SCR, and MBS Traf­fic De­scrip­tors 101 Con­clu­sions 105

Chap­ter 9 Ap­pli­ca­tion of a Call Ad­mis­sion Con­trol (CAC) Al­go­rithm 107

 9.1 Call Ad­mis­sion Con­trol (CAC) 107

 9.2 Scal­ing the PCR and MBS Traf­fic De­scrip­tors 108

 9.3 Ap­pli­ca­tion of CAC 109

 9.4 Com­par­ison of Ef­fec­tive Band­width Ver­sus CLR Tar­get and Queue Length 10

 9.5 Cal­cu­lat­ing Ef­fec­tive Band­width for an Nrt- ​VBR Ser­vice 113

 9.6 Anal­ysis of Re­sults for Nrt- ​VBR 115

 9.7 Max­imal- ​Rate En­ve­lope 116

 9.8 Tol­er­able Queue Length Due to De­lay 117

 9.8.1 Ef­fec­tive Band­width for an Rt- ​VBR Class of Ser­vice 118

 9.8.2 Ef­fec­tive Band­width for an Nrt- ​VBR Class of Ser­vice 122

 9.9 The Im­por­tance of Scal­ing Down the Traf­fic De­scrip­tors 124

 9.10 Anal­ysis of Re­sults of Ta­ble 9- 10 127

 9.11 The Ef­fect of Cell Loss on Voice Qual­ity 128 Con­clu­sions 130

 Chap­ter 10 Queu­ing and Shap­ing 133

 10.1 Sources of De­lay in Voice Net­works 133

 10.2 Re­la­tion­ship Be­tween CLR and Queue De­lay 134

 10.3 De­lay Due to Queu­ing and Shap­ing 135

 10.4 Burst­ing the Buffer Queue 136

 10.5 Buffer Mod­el­ing 13

 10.5.1 Steady State 7

 10.6 Op­net Sim­ula­tion 138

 10.6.1 The M/ M/1 Queu­ing Mod­el 13

 10.7 The Op­net Queu­ing Mod­el 13

 10.8 Op­net Re­sults 139

 10.9 Cal­cu­la­tion of Ex­pect­ed Av­er­age De­lay per Talk­spurt 141

 10.10 Op­net Re­sults on Band­width 145

 10.10.1 Key to Ta­ble 10- 2 6

 10.11 Di­vid­ing Chan­nels in­to More Than One Con­nec­tion 149

 Con­clu­sions

Chap­ter 11 Main Con­clu­sions 153

Appendix and Other

 Ap­pendix A  FRF.11, FRF.8, and FRF.5 157

 Ap­pendix B  Talk­spurt Prob­abil­ity Ta­bles 159

 Ap­pendix C  M/ M/1 Queu­ing Sys­tem 177

 Ap­pendix D Markov's In­equal­ity Al­go­rithml 179

 Ap­pendix E Mar­ket Re­search in­to Voice over Al­ter­nate Net­works 181

 Ap­pendix F Dig­ital Pri­vate Net­work Sig­nal­ing Sub­sys­tem (DP­NSS) 187

 Ap­pendix G QSIG 189

 Ap­pendix H For­mu­las for Ta­bles 191

 Glos­sary 209

 Ref­er­ences 217

 Bib­li­og­ra­phy 225

 In­dex 229

 About the Au­thor

Fore­word

 For years, the trans­mis­sion of voice over the PSTN has been im ple­ment­ed over TDM net­works. Traf­fic en­gi­neer­ing rules for these net­works have been un­der­stood for well over half a cen­tu­ry, but the shift to voice over pack­et net­work­ing in­tro­duces a dif­fer­ent op­er­at ing paradigm. The chal­lenge to net­work op­er­ators is to de­liv­er a Qual­ity of Ser­vice (QoS) That at least match­es the voice qual­ity that TDM net­works have achieved, while re­duc­ing costs through im proved band­width ef­fi­cien­cy and sta­tis­ti­cal gains.

 The TDM codec G.711 is in­ef­fi­cient with re­spect to the use of net­work re­sources. Com­plex voice codecs, such as G.729A/ B, or mo­bile codecs such as AMR, ACELP, or FR- ​GSM can en­code voice sig­nals in­to da­ta units that pro­duce through­put rates at far less than the G.711 rate of 64 Kbps. Fur­ther­more, con­ver­sa­tion­al voice in­volves two states, that of talk­ing and that of lis­ten­ing. When a per­son is lis­ten­ing, there is no need to send dig­itized voice sam­ples in­to a pack­et or cell re­lay net­work to­wards the ac­tive talk­er. This con­cept is known as si lence sup­pres­sion, voice ac­tiv­ity de­tec­tion, or dis­con­tin­uous trans mis­sion and lends well to achiev­ing sta­tis­ti­cal gains in a pack­et net­work.

 The pur­pose of this book is to pro­vide an in­sight in­to some of the chal­lenges and de­ci­sions fac­ing net­work en­gi­neers and plan­ners as they de­ploy voice ser­vices over their pack­et net­works. The ob jec­tive is to ex­plain the dy­nam­ics of con­ver­sa­tion­al speech and to em­pow­er the net­work plan­ner with the req­ui­site knowl­edge to ac cu­rate­ly size the traf­fic de­scrip­tors of mul­ti­plexed voice chan­nels.

 To that end, the au­thor, Juli­et Bates, seeks to de­fine a rep­re­sen tative mod­el of voice traf­fic in or­der to make rec­om­men­da­tions for the ef­fec­tive use of the so­phis­ti­cat­ed traf­fic man­age­ment ca­pa­bili ties of ATM. This book de­scribes how to de­ter­mine the ex­tent to which over­book­ing can be ap­plied in or­der to achieve sta­tis­ti­cal gain of si­lence sup­pressed and com­pressed voice chan­nels, while pro­vid­ing a Cell Loss Ra­tio (CLR) Guar­an­tee.

 Tak­ing this a step fur­ther, it is pro­posed in this book that the de­ter­mi­na­tion of the Peak Cell Rate (PCR), Sus­tain­able Cell Rate (SCR), and Max­imum Burst Size (MBS) ATM traf­fic de­scrip­tors should be based on the math­emat­ical mod­el­ing of con­ver­sa­tion­al speech. Net­works should be, and can be, en­gi­neered such that over­pro vi­sion­ing is not nec­es­sary. ATM mech­anisms for QoS are suit­able for voice ser­vices, de­spite the fact that con­ver­sa­tion­al be­hav­ior is mem­ory­less (sta­tis­ti­cal­ly speak­ing). This book de­scribes how to main­tain QoS for voice traf­fic. ATM al­ready has the req­ui­site QoS ca­pa­bil­ities.

 Some of the prin­ci­pals are ap­pli­ca­ble to IP net­works, but to tru­ly en­gi­neer QoS, ser­vice providers must look to ATM. Read­ers of this book will learn about the op­por­tu­ni­ties for sta­tis­ti­cal gain, what is prac­ti­cal, and what com­pro­mis­es or risks can be tak­en to achieve a high QoS at op­ti mal net­work uti­liza­tion. Bri­an A. Day Prod­uct Line Man­ag­er Voice Gate­ways Broad­band Net­work­ing Di­vi­sion Al­ca­tel Cana­da

 Pref­ace

 Eco­nom­ic gains in voice trans­mis­sion over a net­work are made pos­si­ble by two fac­tors: low bit rate through com­pres­sion, and ex ploita­tion of speech si­lence du­ra­tions. Over­es­ti­ma­tion of band width re­quire­ments will cause un­der­uti­liza­tion of the net­work re­sources and in­creased call block­ing prob­abil­ity. Un­der­es­ti­ma tion may cause cell or pack­et loss.

 The po­ten­tial for sta­tis­ti­cal gain in mul­ti­plexed voice sources is in­hib­it­ed by the in­ter­mit­tent use of sin­gle voice chan­nels for fax and un­com­pressed voice, which may com­pli­cate the nor­mal meth­ods of set­ting traf­fic de­scrip­tors. Over­book­ing can be ap­plied to com­pen­sate. The ex­tent to which over­book­ing can be ap­plied is a sub­ject of this book to­geth­er with cal­cu­la­tions to show the risk of cell loss and the as­so­ci­at­ed de­lay (and de­lay vari­ation) In­curred in the buffer. The Qual­ity of Ser­vice (QoS) Pa­ram­eters, Cell Loss Ra­tio (CLR) And Cell De­lay Vari­ation (CDV), which ap­ply to Asyn­chronous Trans­fer Mode (ATM), Con­stant Bit Rate (CBR), and Vari­able Bit Rate (VBR) Ser­vice class­es are con­trolled by the size of the buffer at the egress port and the ser­vice rate of the queue in the buffer.

 The tar­get CLR will re­flect the prob­abil­ity that the buffer could be ex­ceed­ed. In pro­vi­sion­ing ef­fec­tive band­width, the Call Ad­mis­sion Con­trol (CAC) Al­go­rithm is aware of the QoS tar­gets re­quired by the se­lect­ed ser­vice class. This book de­scribes a method for the re duc­tion of band­width, due to si­lence sup­pres­sion, and the de­ter mi­na­tion of a scal­ing fac­tor which can be ap­plied with a risk of cell loss se­lect­ed in or­der to main­tain the CLR and CDV of the cho­sen ATM ser­vice cat­ego­ry.

 Juli­et Bates Prin­ci­pal Con­sul­tant Pro­fes­sion­al Ser­vices Group Broad­band Net­work Di­vi­sion Al­ca­tel Tele­com UK

 Ac­knowl­edg­ments

 I wish to ac­knowl­edge Brunei Uni­ver­si­ty, Uxbridge, UK for al­low ing me to re­pro­duce ma­te­ri­al from my Ph. D. the­sis (Ju­ly 2001). Al­so, I par­tic­ular­ly thank Dr. Michael Berwick in the De­part ment of Elec­tron­ic and Com­put­er En­gi­neer­ing at Brunei, and Pro fes­sor Mal­colm Irv­ing, who both pro­vid­ed ex­cel­lent sup­port and ad­vice dur­ing my time at Brunei Uni­ver­si­ty. I grate­ful­ly ac­knowl­edge the tech­ni­cal ad­vice I have re­ceived from Mustapha Ais­saoui. I al­so thank Bryan Ed­wards for his en­cour­age­ment, and Tim For­shaw, Dave Hills, and Mike Wilkin­son for kind­ly re­view­ing each draft and pro­vid­ing help­ful com­ments and ad­vice.

 With­in the Ap­pendix­es of this book, there are ex­tracts from com pre­hen­sive re­ports in­to the mar­ket de­vel­op­ment prospects for voice over al­ter­na­tive tech­nolo­gies, and I thank the Yan­kee Group for per­mit­ting this ma­te­ri­al to be in­clud­ed. LT.4.303 In­tro­duc­tion Fa­mil­iar­ity with the fun­da­men­tals of Asyn­chronous Trans­fer Mode (ATM) Net­work­ing is as­sumed. While a very large num­ber of books have been pub­lished on this sub­ject, Bib­li­og­ra­phy re­fer ences [1] and [2] will pro­vide a ground­ing in the ar­eas cov­ered in this work.

 Chap­ter 1 pro­vides an in­tro­duc­tion and de­scribes voice ac­cess mod­els for mo­bile voice net­works, such as the Uni­ver­sal Mo­bile Telecom­mu­ni­ca­tions Sys­tem (UMTS), com­bin­ing the ATM and In ter­net Pro­to­col (IP) In the core.

 Chap­ter 2 de­scribes the voice ac­cess meth­ods and sig­nal­ing pro to­cols em­ployed in fixed voice net­works—for ex­am­ple, where a Pri­vate Branch Ex­change (PBX) Ac­cess­es an ATM core net­work.

 Chap­ter 3 dis­cuss­es re­cent­ly com­plet­ed re­search in voice char ac­ter­iza­tion and de­fends the ra­tio­nale for choos­ing the dis­tribu tion and pa­ram­eters best for a voice mod­el. Al­so in­clud­ed are al­ter­na­tive ways of build­ing a mod­el that will ap­prox­imate the ar ri­val pro­cess us­ing key char­ac­ter­is­tics ob­tained from re­al traf­fic.

 Chap­ter 4 de­scribes the fram­ing for­mat when voice is trans mit­ted over an ATM net­work. Three dif­fer­ent types of ATM Adap tation Lay­er (AAL) Can be used to en­cap­su­late cod­ed voice. This sec­tion helps you as­sess the fram­ing over­heads in­curred by each type of AAL.

 Chap­ter 5 ex­plores var­ious meth­ods of ser­vic­ing the con­tribut ing chan­nels used by each AAL type and cal­cu­lates the corre spond­ing de­lay im­plied by each so­lu­tion. The dif­fi­cul­ties of siz­ing band­width re­quire­ments for com­pressed and si­lence- ​sup­pressed voice chan­nels in the pres­ence of oth­er ap­pli­ca­tions are ex­plained.

 Chap­ter 6 sets out the Cell Loss Ra­tio (CLR) And Cell De­lay Vari­ation (CDV) Per­for­mance tar­gets, which must be achieved to en­sure ad­her­ence to a par­tic­ular ATM Qual­ity of Ser­vice (QoS)

>> Download Op­ti­miz­ing Voice in ATM/ IP Mo­bile I Net­works

Correct and Upload by Anh Minh Dao