QUALITY OF SERVICE AND TRAFFIC PREEMPTION

FOR MULTI-PROTOCOL LABEL SWITCHING / GENERALIZED MULTI-

PROTOCOL LABEL SWITCHING NETWORKS

MOHD NAZRI BIN MOHD WARIP

A project report submitted in partial fulfilment of the

requirements for the award of the degree of

Master of Engineering (Electrical – Electronics & Telecommunication)

Faculty of Electrical Engineering

Universiti Teknologi Malaysia

MAY 2007

iii

ACKNOWLEDGEMENTS

I would like to express my gratitude and appreciation to my supervisor, Prof.

Dr Norsheila Bt Fisal, for his guidance in the execution of the project, for keeping

me on my toes, and for his kind understanding. I am especially grateful for all the

help he provided and resources he made available without which the project would

not have reached its current stage. I would also like to thank Dr Zaharuddin bin

Mohamed, for being most efficient in coordinating the project. My

acknowledgement also goes out to the project presentation assessors, Dr Sharifah

Hafizah Bt Syed Ariffin and Dr Sevia Mahdaliza binti Idrus Sutan Nameh, who have

given me much advice and guidance during the project presentation. Last but not

least, I would like to thank my family for just being there, giving me the strength

and much needed moral support.

iv

ABSTRACT

The Multi-Protocol Label Switching (MPLS) framework is used in internet

service provider (ISP) and as a backbone to Internet Protocol (IP) to provide

guaranteed efficient bandwidth and Quality of Service (QoS) provisioning in the

network. This project seeks to investigate the QoS using DiffServ mechanism over

MPLS and then compare the network performance in legacy networks. As a result,

better network performance is observed with the integration of DiffServ over MPLS.

DiffServ over MPLS provide the capability of the microflow traffic for each class-

type in an aggregated packet stream with a LSP. Traffic preemption and resource

preemption are introduced in QoS over MPLS to demonstrate that the class of traffic

can be classless if MPLS setup-priority and holding-priority are pre-determined at

user defined level.

The architecture that leverages the QoS capabilities of MPLS is simulated by

generating the Class of Traffics Assured Forwarding (AF PHB) and Expedited

Forwarding (EF PHB) comparable with Best Effort (BE) traffic. Then, traffic

preemption is configured to determine priority traffics in MPLS.

DiffServ over GMPLS is used in simulation to show that QoS will improve

the throughput and bandwidth utilization by generating multiple lambda (nλ) and

packet classification with class-based-queuing. Finally, network analysis is

constructed on the impact of all traffics classification in term of end-to-end delay

and jitter

v

ABSTRAK

Kerangka MPLS digunakan pada penyedia perkhidmatan Internet (ISP) dan

merupakan tulang belakang kepada IP untuk menyediakan jaminan kekesan jalur

lebar serta kualiti perkhidmatan (QoS) di dalam rangkaian. Projek ini bertujuan

menyiasat kualiti perkhidmatan mengunakan mekanisme DiffServ melalui MPLS

dan membandingkan prestasinya dengan rangkaian sedia ada. Perbandingan kajian

simulasi di antara topologi rangkaian BE, DiffServ dan MPLS adalah untuk tujuan

rujukan trafik. Hasilnya, pengukuran prestasi rangkaian adalah lebih baik jika

rangkaian digabungkan di antara DiffServ dan MPLS. DiffServ melalui MPLS

berkebolehan menyediakan trafik aliran micro untuk setiap klas trafik di dalam

aliran packet dengan LSP. Traffic Preemption dan resource pre-emption

diperkenalkan di dalam DiffServ melalui MPLS bagi menunjukan bahawa

pengkelasan traffic boleh diubah jika penguna mengkonfigurasikan terlebih dahulu

setup-priority dan holding-priority.

Kemampuan senibina kualiti perkhidmatan (QoS) boleh dijanakan dengan

membandikan simulasikan trafik pengkelasan Assured Forwarding (AF). Expedited

Forwarding (EF) dan Best Effort (BE). Seterusnya, konfigurasi traffic pre-emption

digunakan untuk memnentukan keutamaan trafik di MPLS.

DiffServ melalui GMPLS digunakan di dalam simulasi untuk menunjukkan

bahawa kualiti perkhidmatan akan meningkat dengan kepelbagaian lambda dan

klasifikasi paket dengan CBQ. Akhir sekali, satu analisa rangkaian dijalankan

berdasarkan impak kesemua klasifikasi trafik dari segi kelengahan hujung ke hujung

dan keketaran dan dioah didalam tesis ini.

vi

TABLE OF CONTENT

CHAPTER TITLE PAGE

DECLARATION ii

ACKNOWLEDGEMENTS iii

ABSTRACT iv

ABSTRAK v

TABLE OF CONTENTS vi

LIST OF TABLES x

LIST OF FIGURES xi

LIST OF ABREVIATIONS xii

LIST OF APPENDICES xiii

1 INTRODUCTION 1

1.1 Overview of QoS and MPLS/GMPLS networks 1

1.2 Problem statement 3

1.3 Objectives of the project 3

1.4 Scope of the project 4

1.5 Thesis Structure 4

2 MULTI-PROTOCOL LABEL SWITCHING (MPLS) AND

QUALITY OF SERVICE (QOS) 6

2.1 Multi-Protocol Label Switching (MPLS) 6

2.1.1 Labeling Packets 8

2.1.2 Label Switched Path (LSP) 8

2.1.3 label Distribution Protocol 10

2.2 Generalized Multiprotocol Label Switching (GMPLS) 12

vii

2.3 Quality of Service (QoS) 13

2.4 Differentiated Services (DiffServ) 15

2.4.1 DiffServ Packet Scheduling 15

2.5 Per-Hop Behavior (PHB) 16

2.5.1 Default PHB 17

2.5.2 Class-Selector PHB 17

2.5.3 Expedited Forwarding (EF) PHB 18

2.5.3.1 Definition EF PHB 18

2.5.3.2 Delay and Jitter 22

2.5.3.3 The output buffered model with EF FIFO 23

2.5.4 Assured Forwarding (AF) PHB 23

3 MPLS QOS IMPLEMENTATION 25

3.1 Overview 25

3.2 MPLS Experimental Field 27

3.3 DiffServ of MPLS 28

3.3.1 E-LSP 30

3.3.2 L-LSP 32

3.3.3 DiffServ Tunneling Models over MPLS 34

3.4 Traffic Management 38

4 THE MPLS SIMULATOR USING NS-2 41

4.1 Conceptual model of MNS supporting QoS 41

4.2 Conceptual model of MNS supporting QoS 42

4.3 Design and Implementation of MNS 43

4.3.1 Label Switching 43

4.3.2 MPLS Real-time Traffic Processing 44

4.3.3 Resource Reservation 45

4.3.4 Class Level 46

5 THE QOS AND MPLS MODELING AND SIMULATION 47

5.1 DiffServ over MPLS Network Model 47

5.2 DiffServ over GMPLS Network Model 48

5.3 Project Methodology 49

viii

5.3.1 Design Flow Chart for QoS and Traffic Preemption 50

5.4 Network Topology 52

6 SIMULATION RESULTS AND ANALYSIS 54

6.1 Overview 54

6.2 MPLS and GMPLS Networks Configuration Simulation 54

6.3 Internet Traffic Model 56

6.4 Simulation Results 57

6.4.1 Result 1: Best Effort (BE) – DiffServ Disabled and

MPLS Disabled 57

6.4.2 Result 2: DiffServ Enabled Only and MPLS Disabled 59

6.4.3 Result 3: MPLS without DiffServ 60

6.4.4 Result 4: QoS DiffServ over MPLS 61

6.4.5 Diffserv-aware-MPLS and Traffic Preemption 63

6.4.6 Result 6: Comparison End to End (end-to-end) Delay 64

6.4.7 Result 7: Comparison Delay for EF/AF Traffics in

(G)MPLS per λ 66

6.5 Performance Analysis and Discussion 68

7 CONCLUSIONS AND FURTHER WORK 71

7.1 Conclusion 71

7.2 Potential Proposed Future Works 72

REFERENCES 73

Appendices A – D 78 - 94

ix

LIST OF TABLES

TABLE NO. TITLE PAGE

2.1 IP Precedence-to-DSCP Mapping 15 3.1 QoS Policy 27 3.2 Mandatory PHB Mappings for L-LSPs 32 3.3 Comparing E-LSPs and L-LSPs 33 3.4 DiffServ Tunneling Models over MPLS with PHP 37 3.5 Tunneling Models over MPLS Without PHP 38 6.1 DiffServ over MPLS Bandwidth Utilization 63 6.2 DiffServ over MPLS and Traffic Preemption 64 6.3 QoS Comparison on legacy networks 68 6.4 QoS comparison on Class of Services Traffics 69 6.5 QoS performance comparison 70

x

LIST OF FIGURES

FIGURE NO. TITLE PAGE

2.1 MPLS Concepts 7

2.2 MPLS Label header 8

2.3 DiffServ Hierarchical Packet Scheduler 16

3.1 QoS in 3 Dimensions definition 26

3.2 MPLS Shim Header 29

3.3 MPLS Network Using an E-LSP 31

3.4 MPLS Network Using L-LSPs 33

3.5 MPLS Network Combining L-LSPs and E-LSPs 34

3.6 Pipe Tunnelling Model 36

4.1 MPLS Node – Conceptual Model of MNS v2.0 42

4.2 MPLS Architecture Node for Label Switching 44

5.1 DiffServ over MPLS Network Model 48

5.2 DiffServ over GMPLS Network Model 49

5.3 Project Methodology Flow Chart 51 5.4 MPLS Networks Topology 52

5.5 GMPLS Network Topology 53

6.1 MPLS / GMPLS networks topology simulation 55

6.2 Best Effort (BE) Network Simulation 58

6.3 BE Traffic Simulation at Node 0 58

6.4 DiffServ Enabled Simulation at Node 0 59

6.5 MPLS enabled Simulation at Node 9 60

6.6 DiffServ over MPLS Network Simulation at LSR 1-3-5-7-9 62

6.7 DiffServ over MPLS Simulation at Node 9 63

6.8 DiffServ over MPLS with Traffic Preemption Simulation at

Node 9 64

xi

6.9 MPLS/GMPLS end-to-end delay and jitter measurement 65

6.10 end-to-end delay and jitter performance for DiffServ enabled 65

6.11 end-to-end delay and jitter performance for

DiffServ over MPLS 66

6.12 end-to-end delay performance GMPLS Network Simulation 67

6.13 Comparison end-to-end delay for GMPLS per nλ 67

xii

LIST OF SYMBOLS

B - Packet Size D - Exponential Service Time Distribution E - Latency M - Poisson Arrival Process R - Transfer/Service Rate s - Packet Service Time Q(x) - Function λ - Optical Link ρ - Link Utilization µ - Departure aj - Time Last Bit dj - Time Last Bit Ea - Error Lj - Size Last Bit tq - Mean Time tw - Queuing Delay

xiii

LIST OF ABREVIATIONS

AF Assured Forwarding ATM Asynchronous Transfer Mode FEC Forwarding Equivalence Class BA Behaviour Aggregate BE Best Effort BGP Border Gateway Protocol BW Bandwidth CoS Class of Service CR-LSP Constraint-based Routing Label Switch Path CBQ Class Based Queuing CR-LDP Constraint Route Label Distribution Protocol CBR Constant Bit Rate CDR Commited Data Rate CBS Commited Burst Size CAR Commited Access Rate CBWFQ Class-Based Weighted Fair Queuing CE Customer Edge CPN Customer Premises Network DiffServ Differentiated Services DSCP Differentiated Services Code Point

xiv

DTS Digital Transmission System DRR Deficit Round Robin ER-LSP Explicit Route Label Switch Path EF Expedited Forwarding E-LSP Experimental inferred Per Service Class Label Switch Path EXP Experimental Bit EBS Excess Burst Size ERB Explicit Route Information Base FEC Forwarding Equivalence Class FIFO First In First Out FTP File Transfer Protocol GMPLS Generalized Multi-Protocol Label Switching IP Internet Protocol ISP Internet Service Provider IETF Internet Engineering Task Force IS-IS Intermediate System to Intermediate System ITU International Telecommunication Union LDP Label Distribution Protocol LER Label Edge Router LSP Label Switched Path LSR Label Switch Router L-LSP Label inferred Per Service Class Label Switched Path LC-ATM Logical Connection Asynchronous Transfer Mode LIB Label Information Base LMP Link Management Protocol

xv

LFIB Label Forwarding Information Base LSPID Label Switched Path Identification MPLS Multi-Protocol Label Switching MTU Maximum Transfer Unit MNS MPLS Network Simulator NS-2 Network Simulator Version 2 NCT Network Control Traffic nrt Non Real-Time OSPF Open Short Path First OXC Optical Cross-Connect PHB Per-Hop-Behaviour PDR Peak Data Rate PBS Peak Burst Size PE Provider Edge PHP Penultimate Hop Popping PFT Partial Forwarding Table PSTN Public Switching Telephone Network QoS Quality of Service RED Random Early Detection RFC Request For Comment RSVP Resource ReSerVation Protocol RSVP-TE Resource ReSerVation Protocol Traffic Engineering rt Real-Time S Stack SDH Synchronous Digital Hierarchy

xvi

SONET Synchronous Optical Network SLA Service Level Agreement ServiceID Service Identification SAN Storage Area Network TTL Time To Live TLV Type-Length-Value TE Traffic Engineering TCP Transmission Control Protocol TDM Time Division Multiplexing ToS Type of Service TELNET Remote Terminal Protocol UDP User Datagram Protocol UNI User Network Interface VPI Virtual Path Identifier VCI Virtual Channel Identifier VPN Virtual Private Network VoIP Voice over Internet Protocol VBR Variable Bit Rate WAN Wide Area Network WRED Weighted Random Early Detection

xvii

LIST OF APPENDICES

APPENDIX TITLE PAGE

A Timeline for Project 78

B MPLS/GMPLS Network Source Code in NS-2 79

C QoS over MPLS/GMPLS Source Code in NS-2 86

D QoS and Traffic Preemption for MPLS/GMPLS Source Code

In NS-2 94

2

the best service it can, but there are no controls to preserve higher levels of service

for some flows and not others. This inefficient traffic resource allocation causes

congestion in the network which eventually leads to a drop in network performance.

Over the years, several approaches have been developed to enhance the

internet in order to support the different requirements of different types of traffic and

address the QoS problem. DiffServe is one of the mechanisms able to provide QoS.

In DiffServe, traffics are treated differently based on their QoS requirements.

Thus, in a Best-Effort (BE) environment, what DiffServe does is attempt to provide

better levels of service. However, it is still below par because traffic engineering

remains a problem.

DiffServe and MPLS perform a similar set of functions that can be combined

when both are implemented in a network. Support of DiffServe in MPLS that was

standardized by the IETF can provide QoS guaranteed service, while keeping

network resource utilization at high level [13].

GMPLS is a more general protocol that extends MPLS to provide common control

and traffic engineering.

Hence, this project studies the Quality of Service (QoS) capabilities and

traffic pre-emption for priority fine-tuning in Multi Protocol Label Switching

(MPLS) and extended Generalized Multi Protocol Label Switching (GMPLS).

GMPLS is not just about protocols the existing MPLS protocols are reused

with relatively small extensions and it is not about any particular technology. It can

be applied to many, including TDM, lambda switching, and pre-existing MPLS

devices. GMPLS is all about the general software architecture of a network element

and network applications on top of the protocol [1].

3

1.2 Problem Statement

The increasing population of internet users are demanding for faster, more

efficient and more reliable service network. As stated earlier, Quality of Service

(Qos) is now a prerequisite for the transport of high bandwidth voice and video

services, as well as critical data.

An insight into the traditional Best Effort method provides a transparency for

the requirements of Quality of service (QoS) and the basis for this project.

In Best Effort (BE), traffic congestion often occurs in the path and this leads

to major flaws related to the bandwidth, delay, jitter and packet loss which are the

parameters for QoS in a network. These are the contributing factors to degrading

network performance. Hence, the need for QoS is vital in order to control and

manage the mentioned parameters.

1.3 Objectives of the Project

The main objective of this project is to investigate the QoS capabilities and

traffic pre-emption for priority fine-tuning in MPLS and extended GMPLS using

Differentiated Services (DiffServ) mechanism. In doing so, the following objectives

are to be met:

• To study and compare the Quality of Service (QoS) under different policy

based used in MPLS / GMPLS networks by modelling the traffic pre-

emption.

• To simulate the performance parameter on DiffServ over MPLS / GMPLS

environment using NS-2 software tool.

4

• To compare results with Class of Service (CoS) such as EF, AF and BE on

ER-LSP and CR-LSP in QoS and Traffic Prremption.

• To generate analysis from the results and draw conclusion.

1.4 Scope of the Project

The project will first estimate and identify the provisioning of Quality of

Service (QoS) capabilities on DiffServ-aware-MPLS and DiffServ-aware-GMPLS

based on RFC2474, RFC2475, RFC2597, RFC2598, RFC3140, RFC3270, RFC3471,

RFC3472 and RFC4594 by generating the CBR traffic into different priority levels,

and then apply different event schedulings such as the following;

(a) Per Hob Behaviour (PHB);

1. Expedited Forwarding [EF] or Premium Service

2. Assured Forwarding [AF]

(b) Best-Effort [BE] – default

(c) CBQ Queuing Mechanism at the nodes.

Second, simulate the resource pre-emption defined in Constrained Routing –

Label Distribution Protocol (CR-LDP) on MPLS / GMPLS by generating different

CBR traffic.

Finally, compare and contrast both the variation in bandwidth studies

obtained and then generate analysis before drawing a conclusion.

1.5 Thesis Structure

This report consists of 7 chapters. Chapter 1 is an introduction presenting a

brief overview of QoS and the trend of the current Internet, problem statement as

well as objectives and scope of this project. Chapter 2 describes the QoS and label

5

switching that can be categorized into MPLS and GMPLS, followed by a brief

explanation on DiffServ. Next, Chapter 3 explains the implementation of QoS in

MPLS network.

The use of NS-2 in the simulation of MPLS/GMPLS network is discussed in

Chapter 4. This is then followed by Chapter 5 which looks into project

methodology and simulation studies. Chapter 6 then presents the analysis of the

performance. The QoS parameters investigated for this network model is also

presented and discussed here. Finally, Chapter 7 draws a conclusion based on the

findings and discussions from the preceding chapters and presents several proposals

to improve the performance of MPLS/GMPLS and extensions to the works of this

project.

CHAPTER 1

INTRODUCTION

1.1 Overview of QoS and MPLS/GMPLS Networks

The internet industry is growing rapidly with increasing demands for the

provisioning of new and more advanced services that are able to dynamically react

to changes within the network. In spite of the internet’s evolution from a network

that provided just Best-Effort (BE) transportation to a network capable of providing

a wide range of services, users are seeking for a more efficient and reliable network

with guaranteed quality of service.

Quality of Service (QoS) is in itself a measurement of how much a service

meets and satisfies its users’ needs. Technically, it is a set of qualities related to the

collective behaviour of one or more objects [1]. In a network, QoS is measured via

bandwidth, delay, jitter and packet loss.

In today’s internet industry, Quality of Service (QoS) plays a significant role

in the implementation of multiservice and converged networks. The parameters of

QoS need to be met to support distinct applications such as voice, video, and data,

and multiple network services such as IP, Ethernet, and ATM.

However, bandwidth, delay, jitter and packet loss are the main problems in

traditional Best Effort (BE), and as such BE doesn’t support QoS requirements.

With Best Effort (BE) there is no attempt to differentiate between the traffic flows

that are generated by different hosts. As traffic flow varies, the network provides