implementation and performance evaluation of two … · implementation and performance evaluation...

74
Implementation and Performance Evaluation of two Reliable MAC Layer Multicast Schemes for Wireless Local Area Network Imran Siddique and Waseem Ahmad September 1, 2008 Master’s Thesis in Computing Science, 2*30 ECTS credits Supervisor at CS-UmU: Thomas Nilsson Examiner: Per Lindstr¨ om Ume ˚ a University Department of Computing Science SE-901 87 UME ˚ A SWEDEN

Upload: ngocong

Post on 14-Jul-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

Implementation andPerformance Evaluation of twoReliable MAC Layer Multicast

Schemes for Wireless LocalArea Network

Imran Siddique and Waseem Ahmad

September 1, 2008

Master’s Thesis in Computing Science, 2*30 ECTS creditsSupervisor at CS-UmU: Thomas Nilsson

Examiner: Per Lindstrom

Umea UniversityDepartment of Computing Science

SE-901 87 UMEASWEDEN

Abstract

In the Wireless Local Area Network(WLAN), IEEE 802.11 is the most popular standarddue to its low cost and easy deployment.

However, IEEE 802.11 performs poorly for the multicast applications. Several prob-lems exist for the multicast services in the wireless networks. Different approaches haveproposed to improve the reliability and performance of multicast transmission in IEEE8022.11 networks.

In this thesis, two approaches are used to enhance reliability and performance ofmulticast traffic. EMCD is an algorithm which detects collision by pausing duringtransmission. PREMA is an algorithm which resolves the collision by jamming thechannel.

The simultion results present key parameters which affect the performance. Theresults show enhancements in the multicast services. EMCD performance is directlyproportional to its collision detection ability Collision detection in EMCD is highlydependent on its collision detection interval (TCDI). The comparative study has beendone on simulation results, which shows PREMA achieves very high channel utilizationand success probability by resolving the collisions. PREMA simply outperforms EMCDand IEEE 802.11 in all scenarios and independent of large network.

ii

Acknowledgments

First of all thanks to Almighty Allah (The most Generous and most Beneficent). Weare thankful to our supervisor Dr. Thomas Nilsson, for all his guidance and valuablesuggestions. He inspired us as a teacher and researcher. He taught us to face and handlethe challenges. We wish to become ambitious and committed like him.

We are grateful to the Director of Studies, Dr. Per Lindstrom, for his remarkablecooperation and exceptional management for our studies.

We are thankful to our friends who helped and support us during our study period.Especially we would like to admire Jahanzeb Tipu for his valuable suggestions. Thanksto Asrar and Shahid for their unforgettable company during studies in Umea.

The most important we would like to express our heartiest gratitude to our parentsand family members for their blessings, support and patience.

Imran Siddique, Waseem Ahmad.

iii

iv

Contents

1 Problem Specification 1

1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Goal Of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.3 Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Introduction to IEEE 802.11 3

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.1.1 System Architecture and Service . . . . . . . . . . . . . . . . . . 4

2.1.2 Distributed Coordination Function (DCF) . . . . . . . . . . . . . 5

2.1.3 Collision Avoidance (CA) and Binary Exponential Backoff (BEB) 6

2.1.4 An Example of DCF Operation . . . . . . . . . . . . . . . . . . . 7

3 Limitations of IEEE 802.11 9

3.1 Limitations of DCF Mechanism . . . . . . . . . . . . . . . . . . . . . . . 9

3.1.1 Unfairness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.1.2 Multicast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

4 Different Approaches for Reliable Wireless Multicast 13

4.1 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

4.1.1 Collision Resolution . . . . . . . . . . . . . . . . . . . . . . . . . 14

4.1.2 Collision Detection . . . . . . . . . . . . . . . . . . . . . . . . . . 14

4.1.3 Channel Reservation . . . . . . . . . . . . . . . . . . . . . . . . . 14

5 Prioritized Repeated Elimination Multiple Access 17

5.1 Introduction to PREMA . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

5.1.1 Bursting and Elimination . . . . . . . . . . . . . . . . . . . . . . 17

5.1.2 PREMA for IEEE 802.11 . . . . . . . . . . . . . . . . . . . . . . 18

5.1.3 Optimal Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 19

5.1.4 Simple Scenarios of PREMA . . . . . . . . . . . . . . . . . . . . 19

v

vi CONTENTS

6 Early Multicast Collision Detection 236.1 Introduction to EMCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236.2 How Protocol Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

6.2.1 Vanguard Transmission . . . . . . . . . . . . . . . . . . . . . . . 256.2.2 Carrier Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256.2.3 Phase III - Jamming/Normal Transmission . . . . . . . . . . . . 26

6.3 Collision Detection Interval . . . . . . . . . . . . . . . . . . . . . . . . . 266.4 PHY and MAC Header . . . . . . . . . . . . . . . . . . . . . . . . . . . 276.5 EMCD Behavior in Different Scenarios . . . . . . . . . . . . . . . . . . . 27

6.5.1 Collision between two Multicast Senders . . . . . . . . . . . . . . 276.5.2 Collision between Unicast and Multicast Senders . . . . . . . . . 276.5.3 Undetectable Collisions . . . . . . . . . . . . . . . . . . . . . . . 28

7 Design and Implementation 317.1 Introduction to GloMoSim . . . . . . . . . . . . . . . . . . . . . . . . . . 317.2 Design and Implementation . . . . . . . . . . . . . . . . . . . . . . . . . 32

7.2.1 Implementation of EMCD . . . . . . . . . . . . . . . . . . . . . . 327.2.2 Working of EMCD . . . . . . . . . . . . . . . . . . . . . . . . . . 347.2.3 Implementation of PREMA . . . . . . . . . . . . . . . . . . . . . 367.2.4 Working of PREMA . . . . . . . . . . . . . . . . . . . . . . . . . 37

8 Evaluation 418.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418.2 Performance Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . 418.3 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

8.3.1 Network Architecture . . . . . . . . . . . . . . . . . . . . . . . . 428.3.2 Physical Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 42

8.4 Evaluation of EMCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428.4.1 Throughput . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428.4.2 Collisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

8.5 Evaluation of PREMA . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478.5.1 Effecting Parameter . . . . . . . . . . . . . . . . . . . . . . . . . 47

8.6 Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 488.6.1 Throughput Comparison of Multicast Traffic . . . . . . . . . . . 488.6.2 Throughput Comparison among Multicast and Unicast Stations 49

9 Conclusion and Further Research 519.1 EMCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

9.1.1 Advantages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 519.1.2 Disadvantages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

9.2 PREMA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529.2.1 Advantages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

CONTENTS vii

9.2.2 Disadvantages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529.3 Further Reading and Research . . . . . . . . . . . . . . . . . . . . . . . 52

9.3.1 EMCD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529.3.2 PREMA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

10 Summary 55

A List of Abbreviations 57

viii CONTENTS

List of Figures

2.1 System architecture of WLAN . . . . . . . . . . . . . . . . . . . . . . . . 42.2 The unicast and multicast services . . . . . . . . . . . . . . . . . . . . . 42.3 The basic mechanism for unicast and multicast in DCF. . . . . . . . . . 52.4 A relationship between Inter Frame Spaces . . . . . . . . . . . . . . . . 62.5 An example of DCF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

3.1 Unfairness between uplink and downlink . . . . . . . . . . . . . . . . . . 103.2 Multicast ACK implosion. . . . . . . . . . . . . . . . . . . . . . . . . . . 11

4.1 Hierarchal view of different approaches for reliable multicast [24] . . . . 13

5.1 The basic access mechanism of PREMA . . . . . . . . . . . . . . . . . . 185.2 A flowchart of PREMA . . . . . . . . . . . . . . . . . . . . . . . . . . . 195.3 A collision scenario between two multicast stations . . . . . . . . . . . . 205.4 A collision scenario between multicast and unicast station . . . . . . . . 21

6.1 EMCD flow diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246.2 Early Multicast Collision Detection. . . . . . . . . . . . . . . . . . . . . 246.3 Different vanguard transmissions by three multicast senders . . . . . . . 256.4 Physical data packet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276.5 Collision detection between two multicast senders . . . . . . . . . . . . . 286.6 Collision detection between unicast and multicast senders . . . . . . . . 286.7 Undetectable collision between unicast and multicast senders . . . . . . 286.8 Undetectable collision between two multicast senders . . . . . . . . . . . 29

7.1 Sequence diagram of EMCD with successful transmission . . . . . . . . 347.2 Sequence diagram of EMCD with unsuccessful transmission . . . . . . . 357.3 Sequence diagram of PREMA . . . . . . . . . . . . . . . . . . . . . . . . 38

8.1 EMCD throughput on different TCDI values. . . . . . . . . . . . . . . . 438.2 EMCD throughput on different CW values. . . . . . . . . . . . . . . . . 448.3 Undetection probability at different number of selectables . . . . . . . . 45

ix

x LIST OF FIGURES

8.4 Detection probability at different number of selectables . . . . . . . . . . 468.5 Collision at MAC layer for vanguard transmissions . . . . . . . . . . . . 478.6 Prema results on different values of q and h parameters . . . . . . . . . 488.7 Throughput comparison of PREMA, EMCD and 802.11 . . . . . . . . . 498.8 Throughput comparison of PREMA, EMCD and 802.11. . . . . . . . . . 49

List of Tables

6.1 CIFS relation to Inter Frame Spaces . . . . . . . . . . . . . . . . . . . . 26

7.1 Protocols and Models supported by GloMoSim at each layer . . . . . . . 317.2 Important data members used in GlomoMacEmcd data structure . . . . 327.3 Frame types used in Emcd MacFrameType data structure . . . . . . . . 337.4 Important states used in Emcd MacStates data structure . . . . . . . . 337.5 Important parameters of GlomoMacPrema . . . . . . . . . . . . . . . . . 367.6 Some constant parameters used in implementation of PREMA . . . . . 377.7 Frame types used in implementation of PREMA . . . . . . . . . . . . . 377.8 MAC layer states used in implementation of PREMA . . . . . . . . . . 37

8.1 EMCD parameters used in simulations . . . . . . . . . . . . . . . . . . . 428.2 EMCD TCDI values used in simulation . . . . . . . . . . . . . . . . . . . 43

xi

xii LIST OF TABLES

Chapter 1

Problem Specification

1.1 Background

The Wireless Local Area Network (WLAN) standard, IEEE 802.11 is a popular tech-nology due to its best effort services, low cost and ease of deployment. IEEE 802.11became more and more popular due to its support for multimedia applications. Sincemultimedia applications require more bandwidth, therefore multicast services are usedas a communication method in IEEE 802.11 networks. When a sender is transmittingthe same data packet simultaneously to its neighbor or group of neighbors is known asmulticast [22].

In IEEE 802.11, unicast traffic is protected by the Automatic Repeat Request (ARQ)mechanism which is based on acknowledgment (ACK). But in multicast traffic, the ARQmechanism is not possible because multiple receivers try to send their ACKs at the sametime to one sender which causes a feedback implosion. Packets losses due to overlappingcells and link adaptation are also considered as a problem creating issues in multicasttraffic. Therefore different protocols are proposed to solve the prediscussed issues [24].

1.2 Goal Of Thesis

The main task of this thesis is to implement two protocols, Early Multicast Collision De-tection (EMCD) [22] and Prioritized Repeated Eliminations Multiple Access (PREMA)[12] in IEEE 802.11. These protocols should be implemented in GloMoSim (Global Mo-bile Information System Simulator), written in the C language. The master thesis is acomparative study of these protocols aiming to make them more flexible for multicast.

EMCD is an algorithm, designed for IEEE 802.11 networks. The objective is toimplement EMCD for multicast in GloMoSim. A multicast sender performs the CCA(clear channel assessment) during its early pause in the transmission before sending thepacket. If the channel is busy, the station aborts the transmission after a fixed timeinterval and schedules a retransmission.

PREMA is a protocol that borrows the idea of bursting from EY-NPMA [5]. Thebursting mechanism enhances the performance in unicast as well as in broadcast net-works. PREMA was intended for a wireless broadcast network with few or many nodes[12]. PREMA consists of sensing the channel and prioritizing the nodes for sending

2 Chapter 1. Problem Specification

data by the elimination process, which is based on different priority schemes. It makesPREMA more flexible and adaptable due to its quality of service (QoS), reliability andhigh performance achievement for the multicast transmission in multimedia applications.

The goal of the thesis is to increase the reliability for multicast transmission, fordifferent scenarios in PREMA. The objective is to implement PREMA by using backoffschemes and concept of bursting in order to achieve maximum throughput. The finaltask is to evaluate the performance of EMCD and PREMA by taking real time scenarios.

1.3 Tasks

The thesis work comprises of 40 points (20 x 2) and is divided as follow:

– An in depth study about the multicast problems in IEEE 802.11 networks.

– A comprehensive study of EMCD and PREMA.

– A study of the design of GloMoSim.

– Design and implementation of EMCD and PREMA.

– Testing the implementation.

– Design and simulate different scenarios to analyze the results.

Chapter 2

Introduction to IEEE 802.11

2.1 Introduction

Wireless networks are nowadays widely used and experienced a great success after thedevelopment of Internet. There are two types of networks used: centralized and dis-tributed. A centralized network is centrally controlled by the access point (AP). Adistributed network has no central point, a wireless station accesses the network us-ing a access mechanism. Several standards exists for WLAN like IEEE 802.11 [4] andHiperLAN (High Performance Radio LAN)1 /and 2 [5].

IEEE 802.11 is also know as WiFi. It is the most popular and widely used standarddue to its simplicity and low cost. HiperLAN/1 and 2 are not well known standards dueto their complexities and are not considered to be a part of our study.

In 1997, Institute of Electrical and Electronics Engineers (IEEE) released the 802.11standard that also defines the Media Access Control (MAC) and physical (PHY) layerspecification. There are three different types of PHY layer specifications described,Frequency Hopping Spread Spectrum (FHSS), Direct Sequence Spread Spectrum (DSSS)and Infra Red (IR). The FHSS and DSSS operate on the 2.4 GHz ISM (Industrial,Scientific and Medical) band, which is license free. FHSS and DSSS have a maximumdata transfer rate of 2 Mbps.

After 2 years, IEEE enhanced the physical layer specifications and introduced twonew versions 802.11a [6] and 802.11b [7]. The 802.11b improves the DSSS physical layer,which operates on the 2.4 GHz, and archives the maximum data transmission rate of 11Mbps. The 802.11a OFDM physical layer specification operates on the 5 GHz band withmaximum data transmission rate up to 54 Mbps. However the MAC layer specificationremains the same except for a few parameters which are dependent on the PHY layer.

The MAC basically controls the access of a transmission on the medium. TheMAC protocols are based on two different access mechanisms, the Distributed Coor-dination Function (DCF) and the Point Coordination Function (PCF). PCF is a pollingbased technique which centrally controls the channel and grants access based on polling.Whereas DCF is a multiple access technique based on the Carrier Sense Multiple Accesswith Collision Avoidance (CSMA/CA) mechanism.

4 Chapter 2. Introduction to IEEE 802.11

2.1.1 System Architecture and Service

IEEE 802.11 defines two different type of architectures called Basic Service Set (BSS)and Independent Basic Service Set (IBSS). The BSS infrastructure is formed when one ormore wireless stations are associated with an AP. Several BSS are connected through aDistribution System (DS) that forms an Extended Service Set (ESS) as shown in fig 2.1.All the communication takes place through the AP regardless of the destination address.That means, stations can not communicate with each other directly. In contrast, stationscan communicate with each other directly in an IBSS infrastructure, if they are withinthe range of each other. This allows the formation of a wireless ad-hoc network in theabsence of any network infrastructure.

Figure 2.1: System architecture of WLAN

IEEE 802.11 supports both unicast and multicast services. When a station sends apacket to a single destination is known as unicast and a single sender who transmits apacket to a group of receivers called ”group members”, as shown in fig. 2.2, is knownas multicast.

Figure 2.2: The unicast and multicast services

2.1. Introduction 5

Multicast is an important feature in WLAN, since it is used in several applicationsincluding audio and video streaming, multimedia conferencing, shared white boards,distance learning applications, multi player games etc [15].

IEEE 802.11 treats both services differently. There is no Automatic Repeat Request(ARQ) mechanism for multicast as it is used for retransmission in unicast. Thereforethe IEEE 802.11 is not a reliable protocol for multicast.

2.1.2 Distributed Coordination Function (DCF)

DCF is the basic access mechanism for the IEEE 802.11 networks with Carrier SenseMultiple Access (CSMA). CSMA works as listen before talk scheme, that means thestation senses the medium for a specific time interval called DIFS (DCF Inter FrameSpace) before the transmission [11]. If the medium is idle then the transmission proceeds.Otherwise the station defers the access and waits until the medium becomes idle again,for a DIFS time period.

Figure 2.3: The basic mechanism for unicast and multicast in DCF.

The unicast transmission is followed by an acknowledgment (ACK) and a retransmis-sion technique to make sure that the packets have been received successfully. If the APreceives a unicast packet successfully, then it transmits an ACK after waiting for a shorttime period called Short Inter Frame Space (SIFS), as shown in fig 2.3. If there is noACK received until the ACK timeout, the sender realizes that the packet has been lostand it schedules a retransmission. In the multicast transmission, the ACK and retrans-mission technique can not be applied due to undefined number of receivers as shown infig 3.2. The sender can not receive the multiple ACKs at the same time. If there is noACK then the collision can not be detected and there will be no retransmission.

In the IEEE 802.11, all frame types do not have the same priority, therefore, InterFrame Space (IFS) time intervals are defined [9]. It gives a priority access to the channelbetween the transmissions. The relation between inter frame spaces are shown in fig2.4. There are three types of IFS defined in IEEE 802.11, as described below.

– Short inter frame space (SIFS): has the highest priority and shortest time intervalwhich comes between the packet and the ACK frame. It prevents the other stationsto transmit while a sender is waiting for the ACK.

– PCF inter frame space (PIFS): is shorter than the DIFS time interval used by thePCF, an optional mechanism, used in IEEE 802.11. The AP is centrally controllingthe channel through polling of individual stations. In PCF, the AP waits for a

6 Chapter 2. Introduction to IEEE 802.11

PIFS time period which gives a priority access over the ordinary station, sincethey have to wait a longer time, at least for a DIFS time period.

– DCF inter frame space (DIFS) : has the lowest priority and the largest time intervalthat comes prior to the packet transmission.

Figure 2.4: A relationship between Inter Frame Spaces

The lengths of the IFSs are dependent on the physical layer specification and are mea-sured in time slots. In the DSSS, the SIFS time period is equal to a half slot time andPIFS includes SIFS plus one slot time. The DIFS contains a SIFS plus two time slotstime.

2.1.3 Collision Avoidance (CA) and Binary Exponential Backoff(BEB)

In the CSMA, if two or more stations are trying to access the medium at the sametime that leads to a collision. To prevent this situation, a Collision Avoidance (CA)mechanism works along the CSMA and this is known as the CSMA/CA.

In the WLAN, collision avoidance is used instead of Collision Detection (CD) which isused in Local Area Network (LAN) , e.g. IEEE 802.3 Ethernet [3]. Due to this, wirelessnetworks are not capable of detecting the collisions. In a wired network, stations arecapable of sending and receiving the packet at the same time [19], therefore, it is possibleto detect the collisions. While in the wireless networks the stations can not receive asignal while transmitting, that is the basic characteristics of the wireless communication.The strength of the signal also decreases while it propagates. The other factors likeinterferences, noises and fading also affect the signal strength which makes it difficultfor the sender to detect other signals in the presence of their own signal [19].

CSMA/CA relies on the Binary Exponential Backoff (BEB) algorithm to preventthe collisions. Stations have to wait for an extra contention time period after waitingthe DIFS time period prior to the transmission. The station waits until the mediumbecomes idle for at least DIFS time period and selects a random value called backoff(BO) from a uniform interval [0, CW], where CW is the Contention Window. Thestation starts decreasing its backoff value by one for each time slot. A station with thesmallest backoff value counts down to zero first, wins the access of the medium andstarts the transmission. While other stations pause their remaining backoff value andwait until the medium becomes idle again for the DIFS time period. As the mediumbecomes idle for a DIFS, stations start deceasing their backoffs again where they paused,while a new station chooses a new backoff.

When the transmission starts, the CW is set to be minimum value i.e.CWmin. Aftereach collision the size of the CW becomes double , until it reaches its maximum value,i.e. CWmax. Since the CW size increases exponentially, therefore this algorithm isknown as the Binary Exponential Backoff (BEB). By doubling the CW, increases the

2.1. Introduction 7

probability to select a larger backoff value at the same time and decreases the probabilityof further collisions.

The values of CWmin and CWmax are dependent on the physical layer specifications.In IEEE 802.11b, the possible values of the CW are 15, 31, 63, 127, 255, 511 and 1023.A retransmit limit is defined, that means a frame can only be retransmitted to a limitednumber of times. If the retry limit reaches the maximum number of retries, the framewill dropped and the retry limit will be reset.

After each successful transmission the CW is reset to the CWmin. If a sender hasanother frame to transmit then it selects a new backoff value from the reseted CWwhich is known as the post backoff. This ensures that there is at least a backoff intervalbetween two consecutive transmissions.

The CSMA/CA, reduces the risk of collisions but the collisions may still occur, ifthe backoff of two or more stations reaches zero at the same time or if two or morestations select the same backoff value. There is an ACK and retransmission mechanismfor unicast packets and stations have the chance to retransmits. On the other hand,the condition is worse in the case of multicast. There is no ACK for a multicast packet,therefore we can not detect the collision and the packet is lost. Due to this the recoverymechanism can not be imposed for multicast packets. The CW for a multicast stationis always set to the CWmin and doubling the CW is not possible, that also increasesthe risk of collisions.

2.1.4 An Example of DCF Operation

Fig. 2.5 shows the operation of DCF for both the unicast and the multicast ser-vices. There are three stations, two multicast and one unicast, contending to accessthe medium. The packet of a multicast station (MC1) arrives at the MAC layer firstand the station transmits after waiting the DIFS time period. Since there is no otherpacket in the queue, the backoff procedure is not used. Later on the packets of MC2and unicast station (UC) arrive and the stations sense the medium but the medium isbusy. They defer access and wait until the medium becomes idle. After the end of thetransmission, the MC1 will not wait for an ACK. Because there is no ACK for multicastservice, see section 2.1.2. All three stations wait for the DIFS and then select the backoffvalues. The MC1, MC2 and UC1 select the backoff values 11, 4 and 7 respectively and

Figure 2.5: An example of DCF

8 Chapter 2. Introduction to IEEE 802.11

start decreasing by one for each time slot. The MC2 has the shortest backoff value andtherefore it reaches the backoff value zero first and it accesses the channel. The MC1and UC1 pause their backoffs and wait for the medium to become idle again.

Again all stations wait for the DIFS time period and MC2 selects a new backoff valuethat is 3. The MC1 and UC1 resume the count down of their backoff counters. Theremaining backoff of UC1 and MC3 is the same, therefore they start transmission at thesame time and it leads to a collision. The UC1 will wait until the ACK timeout, themulticast stations will immediately wait for the DIFS after the transmission is finished.They get access of the medium a bit early than the UC1 because they do not haveto wait for the ACK timeout. The collision is not detectable for the multicast senderand the packet of MC2 is dropped and also the CW remains the same. After the ACKtimeout the UC1 doubles its CW i.e. 63 and selects a new backoff value after waitingthe DIFS. When UC1 get the access to the medium, it will retransmit the same packet,as before.

Chapter 3

Limitations of IEEE 802.11

IEEE 802.11 is a set of standards developed for WLAN. The different standards of IEEE802.11 focus on the different aspects to improve the overall WLAN performance. Forexample 802.11e deals with QoS, 802.11i deals with security issues and etc [1]. In short,all these standards are dealing with several issues to overcome the problems. There arelot of limitations of these standards which are needed to be solved but are not includedin our study.

This chapter presents an overview of some limitations related to the multicast prob-lems, which prevents the IEEE 802.11 standard to deal fairly with multicast traffic asit does with unicast traffic.

3.1 Limitations of DCF Mechanism

IEEE 802.11 has two basic access mechanisms as discussed earlier. DCF has at leastthe following limitations.

– If multiple stations start their communication at the same time, many collisionsoccur, which lower the achieved bandwidth.

– It treats all kind of data traffic in the same way and has no priority mechanism.This leads to the lacking of QoS in the IEEE 802.11 standard.

– DCF serves all stations in the same fashion and makes no distinguish between Ac-cess Point (AP) and a normal station that leads to the unfairness in the downlink.

– DCF is not suitable for multicast traffic due to the lackness of feedback or ACKimplosion.

3.1.1 Unfairness

Since the AP and normal stations both are treated in the same way causes an unbalancedcontention between uplink and downlink traffic. This is one of the major weakness foundin the DCF mechanism in IEEE 802.11 standard. Unfairness in the contention betweenuplink and downlink stations increases with the increase of load of data traffic andbecomes more evident when the load is exceeded from the maximum capacity of the

9

10 Chapter 3. Limitations of IEEE 802.11

system. There are 10 stations of packet size 1024 with the increasing load of 0 till 4Mbps, used in the simulation as shown in fig 3.1. Where the maximum system capacityis 2 Mbps. The fig 3.1 shows uplink and downlink are attaining 100% fairness in thechannel utilization for a certain period (until the load is approximately 1.4 Mbps), butafter some time the downlink decreases dramatically and uplink traffic achieving morenetwork resources approximately upto the desired level.

Figure 3.1: Unfairness between uplink and downlink

3.1.2 Multicast

Transmission of a single packet simultaneously to a station or group of stations is knownas multicast. The IEEE 802.11 networks are based on the collision avoidance methodsand are protected by the following mechanisms for unicast transmissions: 1)ARQ mech-anism 2)backoff mechanism. The IEEE 802.11 network uses the ARQ (stop and wait)protocol for attaining the feedback to ensure reliable transmissions. But it is not possiblefor multicast traffic. Fig 3.2 illustrates the ACK problem in the multicast transmission.The factors that create problems in multicast are stated below.

1. Automatic Repeat Request (ARQ) is an error control method based on a feedbackmechanism to indicate reliable receiving. In this method, additional check bits areused to ensure the reliable reception of a packet and request for retransmission oflost packet. ARQ is not feasible for the multicast traffic due to the ACK implosion.

2. In real time scenarios, both unicast and multicast traffic categories may exist inthe same environment [22]. On collision, both have totally different ways to dealwith the problem due to the ARQ protocol. In unicast, each cw is doubled onthe collision occurrence but it is not feasible for the multicast case. Therefore,unfairness between unicast and multicast traffic arises.

3.1. Limitations of DCF Mechanism 11

3. When multicast streams are distributed in one ESS, the problem arises whenmultiple multicast streams share the same channel. In the ESS, multiple packetlosses are due to the collisions in overlapping cells on the same channel.

Figure 3.2: Multicast ACK implosion.

STA1 sends data packet to multiple receivers AP1, AP2, and AP3 as shown in thefig 3.2 and wait for ACK. The AP1, AP2 and AP3 receive packet and reply with ACKsat the same time which leads to the collision is called ACK implosion.

The main problems for multicast traffic originate from a common source, poor reliablecapability due to the failure in detection of lost multicast packets. Generally, lost packetsor collisions are detected by the ACK or on the feedback, but it is not possible in themulticast as illustrated in fig 3.2. There are different solutions proposed to solve theissue of feedback [2]. When feedback problem is solved, new challenges arises that arelisted below briefly.

1. Which approach is used for ACK (leader or token based approach)?

2. If it is leader based approach, then who is responsible for sending the ACK?

3. If one station does not receive the packet then what will happen, retransmissionor ignore?

Since retransmissions are not possible therefore contention window can not be adjusteddynamically.

12 Chapter 3. Limitations of IEEE 802.11

Chapter 4

Different Approaches forReliable Wireless Multicast

4.1 Related Work

In this chapter, different approaches are discussed to solve the problems with multicasttransmission. The algorithms are categorized according to their functionality and shownin the fig 4.1.

Figure 4.1: Hierarchal view of different approaches for reliable multicast [24]

14 Chapter 4. Different Approaches for Reliable Wireless Multicast

The main categories are briefly described below.

4.1.1 Collision Resolution

The collision resolution approach is based on certain attributes, taken from the Hiper-LAN (High Performance Radio LAN) European standard to support the IEEE 802.11for multicast traffic. The main objective of this technique is to resolve the collisionsbefore the transmission. This is a hybrid approach because it uses certain beneficialproperties of parallel independent geometric distribution as explained in [12] and jam-ming for channel reservation. EY-NMPA[5], PREMA[12] and SEMA[23] are examplesof this approach. Neither of these protocols are used for the multicast in the IEEE802.11 networks. But, these protocols can be used to improve the multicast for IEEE802.11 networks.

4.1.2 Collision Detection

Collision detection approach can be implemented by adapting different techniques. Ba-sically, this approach is based on the feedback. Where the feedback can be achievedduring the transmission or after the transmission. So, collision detection approach canbe divided into two categories: 1) Carrier sensing during transmission and 2) Carriersensing after transmission (ARQ based protocols).Lo and Mouftah [25], ROM[18] and EMCD are the examples for the collision detectionduring transmission. In this approach, a sender detects collisions by pausing during thetransmission and senses the medium for a fixed time interval to detect other potentialtransmitting stations. If the sender detects the collision, it continues the transmissionfor a predetermined interval which is known as the collision detection interval or thresh-old time. The sender does not abort the transmission immediately because to inform theother senders about the collision. If the channel is sensed idle during the threshold timethen sender continues with its normal transmission. Collisions can also be detectedafter transmission. ARQ based protocols are based on this technique to detect collisions.The protocols based on this technique modified the IEEE 802.11 MAC scheme. Kuriand Kusera[13] proposed a protocol, which is an example for this technique.

4.1.3 Channel Reservation

In this approach, a station avoid other stations from accessing the channel until it finishesthe transmission. The channel is reserved by different techniques depending upon thechannel in which it is operated. For example, In black burst[20] the reservation messageis a burst. The winner of the channel would be that station who will have the longestburst. The burst length is defined by the waiting time. So in a same way, the HiperLANbursts are used for reserving the channel. On the other hand, in IEEE 802.11, RTS/CTSexchange packets are used for the reservation of a channel in the unicast transmission.But in multicast, multiple receivers are sending multiple CTS at the same time thatcauses feedback implosion. Therefore the channel reservation is applicable by followingdifferent techniques for the feedback procedure.

Single CTS approach

1. Robust multicast [14],Simple Leader Base Protocol (SLBP) [2] and Beacon drivenLeader Based Protocol (BLBP) [2] select one leader to send one CTS for the whole

4.1. Related Work 15

group. The selection of the leader is a big problem in these protocols.

2. The BMW [21] ensures the transmission reliability by unicast the RTS/CTS ex-change packets with its neighbors on a round robin fashion. The BMW protocolbecomes worse in the large networks.

Multiple CTS approach

Kuri and Kusera [13] proposed an intolerant approach to enable feedback feature inmulticast traffic. A designated station transmits a positive ACK and if any station whohas not received the multicast frame in that multicast group but detects the ACK willsend a NACK (negative ACK). Due to this, it is a fault intolerant and is not feasibletechnique with the large networks.

16 Chapter 4. Different Approaches for Reliable Wireless Multicast

Chapter 5

Prioritized RepeatedElimination Multiple Access

5.1 Introduction to PREMA

PREMA is a collision resolution algorithm which resolves the channel conflicts prior tothe actual packet transmission. PREMA is a simple bursting protocol that can be usedfor multicast transmissions.

Elimination Yield Non-pre-emptive Prioritized Multiple Access (EY-NPMA) [5], wasprobably the first bursting protocol specified in HiperLAN/1. EY-NPMA is quite similarwith the PREMA. EY-NPMA operates in three phases, 1- Prioritization phase, stationswait according to their priority before the transmission. 2- Bursting phase, stationstransmit their bursts followed by a verification slot. Only that stations having longerbursts, survive and enter into the next phase. 3- Yield phase, the remaining stationsperform yield (backoff). The station with the shortest yield time wins the channel andstarts the transmission. PREMA operates with only bursting and elimination phase andrepeat this phase to a certain times. PREMA starts by transmitting a burst. The burstlength is sampled from the geometric distribution with a probability q. The stationwith the longest burst survives by sensing the medium idle and enters into the secondelimination phase. The winner of h consecutive elimination phases finally accesses themedium.

5.1.1 Bursting and Elimination

PREMA is a simple bursting protocol that resolves the collisions between the multicaststations by using elimination phases. The stations are capable of performing operations,jamming the channel and the carrier sensing.

In PREMA, a station transmits a burst rather than sending the actual packet. Theburst contains no data and is used to transmit the noise on the channel, known aschannel jamming. The elimination phase determines which station survives for the nextphase by performing the carrier sensing. The station who have the longest burst andsenses the medium idle survives while other stations defer their access.

All stations sense the medium for at least DIFS time period before starting thetransmission. If the medium is idle, the station starts transmitting the burst with acertain length sampled by the geometric distribution with a probability q. After the

18 Chapter 5. Prioritized Repeated Elimination Multiple Access

burst transmission, the station performs carrier sensing. If the medium is idle for atleast one time slot that means the station have the longest burst and enters into thesecond phase. If the medium is not idle for a time slot the station realizes that the otherstations have longer bursts. Then the station defer access and is eliminated from thebursting contention as shown in the fig 5.1.

Figure 5.1: The basic access mechanism of PREMA

The remaining stations enter into the second phase. The elimination phase continuesuntil the hth elimination. After the elimination phases there is large probability thatthere will be only a single winner.

5.1.2 PREMA for IEEE 802.11

PREMA is not implemented for the IEEE 802.11 standard. We have modified PREMAto work along with this standard. The backoff mechanism is introduced before the elim-ination phase. The multicast stations use the PREMA algorithm and unicast stationsare treated in a same way as in 802.11. The fig. 5.2 describes the operation of PREMA.

5.1. Introduction to PREMA 19

Figure 5.2: A flowchart of PREMA

5.1.3 Optimal Parameters

The performance of PREMA is dependent on the h and q parameters. Where h isa PREMA threshold (number of elimination phases) and q is the probability used inthe geometric distribution to sampled the burst length. The long burst and severalnumber of elimination phases causes a waste of channel utilization and efficiency of thealgorithm. The values of h and q should be optimized. The effect of these parametersare shown in the Section 8.5. The parameter are described detailedly in [12].

5.1.4 Simple Scenarios of PREMA

The behavior of PREMA is different in the different scenarios. Some of the scenariosare discuss below.

20 Chapter 5. Prioritized Repeated Elimination Multiple Access

Collisions Resolution Between Multicast Senders

The procedure starts in the same way as in the DCF. Fig. 5.3 illustrates an examplescenario where six multicast stations are contending for the medium.

Figure 5.3: A collision scenario between two multicast stations

Every station selects a random backoff value from the CW. The stations STA1, STA2,STA3, STA4, STA5 and STA6 select the backoff values 3, 3, 3, 6, 3 and 9 respectivelyand start to count down the values. The possibility of selecting the same BO value ishigh because of the CW size, which is small. The stations STA1, STA2, STA3 and STA5select the same BO value, therefore they enter into elimination phase. In DCF, thesestations, with the same backoff values, lead to the collision. These stations sample aburst length from the geometric distribution with parameter q and start transmittingthe burst. STA4 and STA6 pause their backoffs and wait until the medium becomesidle. STA2 and STA3 sample a short burst and after sensing the medium busy, theydefer their access. STA1 and STA4 sense the medium idle after transmitting the longestburst and then enter into the second elimination phase.

In the second elimination phase only STA5 survives and enters into third eliminationphase, while STA1 is eliminated because of the shorter burst. After the fourth elimina-tion phase, the only survived station is STA5 who wins the access of the medium andstarts the transmission.

Collisions Resolution Between Unicast and Multicast Senders

In this case, unicast stations are treated in the same way as in the DCF. There is nochange in unicast technique. Fig. 5.4 shows the collision scenario between multicast andunicast stations.

5.1. Introduction to PREMA 21

Figure 5.4: A collision scenario between multicast and unicast station

If both, unicast and multicast stations start transmission at the same time that leadsto a collision. It is only possible if the backoff value of two or more stations reaches zeroat the same time. The unicast stations start transmitting the packet while multicaststarts transmitting the burst. At the end of first elimination phase, the multicast stationrealizes the collision and defers access. Unicast station continues with transmission andwaits until ACK timeout. If the frame is collided then the station schedules for theretransmission of the frame.

Since the multicast station transmits burst to contend for the medium, it saves theactual packet and attempts retransmission on failure. In normal DCF, the multicastpacket is lost and there is no retransmission for multicast packets. There is a smalldelay before the actual packet transmission due to elimination phase but it is almost 99%guaranteed that there is only one winner after the elimination phases [12]. Therefore, itis a trade off between the delay and the reliability of the transmission.

22 Chapter 5. Prioritized Repeated Elimination Multiple Access

Chapter 6

Early Multicast CollisionDetection

6.1 Introduction to EMCD

Early multicast collision detection (EMCD) is an algorithm used by multicast sendersto detect collisions among multicast and unicast senders. EMCD is based on the Rom[18] algorithm. EMCD detects collisions between multicast senders in the IEEE 802.11network and also enables retransmission of the lost packets. Therefore, retransmissionincreases the reliability and reduces overhead associated to collision. Collision detectiontechnique used in EMCD is effective in two ways: first, retransmissions is possible forthe collided packet in multicast. Secondly adjusting the contention window (cw) forcollided packet is also decreasing the number of collisions. The collision detection inEMCD makes it more adaptable to introduce different CW schemes to achieve higherthroughput.

Hence, EMCD is based on Rom algorithm differs from other collision detectionprotocols due to its collision detection mechanism. In EMCD, a multicast station detectspotential transmitters during the transmission by introducing a pause and sensing thechannel as opposed to the slotted CSMA which lost packets after the collision. Thesender who detects a collision, continues the transmission for a predetermined intervalreferred to the collision detection interval (CDI) or threshold time rather than abort thetransmission immediately. If the channel is sensed idle during the threshold time thenthe sender continues with the normal transmission.

Lo and Mouftah [25] and Rom[18] proposed similar collision detection algorithmswith a difference of choosing sub intervals. In the Rom algorithm, the pausing subintervals is chosen from a uniform distribution. On the other hand, Lo and Mouftahassumes a non slotted CSMA with the same pausing sub interval for all senders.

6.2 How Protocol Works

The structure of the EMCD algorithm is shown in the fig 6.1.

Figure 6.1: EMCD flow diagram

Figure 6.2: Early Multicast Collision Detection.

EMCD is divided into three phases as shown in the fig 6.2. The first phase is vanguardtransmission Tv, the second phase is the carrier sensing phase and the third phase isjamming or main transmissionm.

Fig 6.3 illustrates that, how EMCD algorithm works for different senders. Thereare three APs (AP1, AP2 and AP3), contending for transmission after the DIFS timeperiod as shown in the fig 6.3. All senders select different Tv transmissions dependupon the division of the packet, which is explained in the next section. Fig 6.3 describes

6.2. How Protocol Works 25

that how multiple stations choose different Tvs and then during the pause, listen for thechannel status which causes the Tm or jamming of the channel.

Figure 6.3: Different vanguard transmissions by three multicast senders

6.2.1 Vanguard Transmission

The basic idea is to divide the original packet into fragments: Vanguard trans-mission (Tv) and jamming or main transmission (Tm) packets. The size of Tv can becalculated from the given equation (6.1) [22].

Tv = Txmin + nT4, (6.1)

n ∈ U{0, (TCDI − Txmin)/T4} (6.2)

Txmin is minimum allowable transmission time for the Tv. If transmission is Txmin,then it means Tv consists of (PHY + MAC) headers without any data. The PHY andMAC headers are described in the section 6.4. Where ’n’ is the number of slots chosenrandomly as shown in equation (6.2). T4 is the smallest time difference that can bedetectable during the two Tvs, transmitted by two different senders at the same time.TCDI is the maximum allowable transmission time of the Tv plus Collision Inter FrameSpace (CIFS). The parameter TCDI is explained in more detail in the section 6.3. Thesize of the Tv depends on the TCDI time.

6.2.2 Carrier Sensing

In the second phase, CIFS starts after the Tv. During the CIFS interval, theclear channel assessment (CCA) operation is initiated to listen for other potential trans-mitters. Then in the next phase, the decision of jamming or actual transmission comes,which is based on the outcome of the CCA operation. If the medium is idle, the senderscontinue with Tm otherwise jamming will be performed for a specific time.

Carrier sensing is a method by which the station can determine the status of thechannel by CCA operation. Physical layer performs the CCA operation and informs theMAC layer about the status of the channel. There are requirements specified in 802.11aWLAN for the CCA to detect the presence of other signals on the same channel.

26 Chapter 6. Early Multicast Collision Detection

The CIFS lower bound is defined by the time required to perform a correct CCA[6]. Generally, lower bound value ¡ 4µs is used for correct CCA operation. Thecorrect CCA requires signal propagation time and the time required to switch fromtransmitting to receiving and vice versa. These terms (signal propagation and re-ceive/transmit turnaround time) are hardware and physical configuration dependentrespectively. Therefore, the CCA requires time to perform correct operation.

The CIFS upper bound is defined according to SIFS, PIFS and DIFS time intervals.The IEEE 802.11a standard [6] defines the relation between Inter Frame Spaces, as listedin table 6.1. CIFS upper bound value must be less than other values. Overall CIFSvalue can be [3,4,5,...,11] to perform correct CCA. The suggested value for CIFS is 10µsin this algorithm.

Inter Frame Space DurationCIFS 10µsSIFS 16µsPIFS 25µsDIFS 34µs

Table 6.1: CIFS relation to Inter Frame Spaces [8].

6.2.3 Phase III - Jamming/Normal Transmission

During CIFS, if the medium is idle then the Tm is started by creating the packetaccording the following equation.Tm = pkt - Tv; (6.3)pkt is the actual packet comes from network layer to the MAC layer.But if the medium is busy during the CIFS interval then the collided packet is createdby the following equation and is transmitted.

Collided packet = TCDI - (Tv+CIFS); (6.4)

Collided packet jams the channel for a certain time period as calculated in equa-tion (6.4). This jamming is actually based on the concept of acknowledging to everyother station about the collision occurrence. After collided packet transmission, theretransmission is done by following the normal DCF procedure.

6.3 Collision Detection Interval

Collision Detection Interval (TCDI) is important for this algorithm as shownin the fig 6.2. TCDI basically affects the performance of EMCD in three dimensions.First, the length of the TCDI determines how many unique transmissions are selectableby the multicast senders. Smaller value of TCDI offers limited selectable unique trans-missions, that increases the probability of collisions by selecting the same transmissiontime. Secondly, greater TCDI value affects the collision detection ability with unicasttransmission. But Tv also causes additional delay and more overhead.

Fig 6.7 shows the possibility of the collision is undetectable by the longer Tv. Cer-tainly, it also affects on delay and overhead if retransmission is scheduled in longer Tv.

6.4. PHY and MAC Header 27

Therefore, these factors demand for optimal threshold value investigation, that wouldbe a better choice for the unique Tvs and it also does not cause longer Tvs.

6.4 PHY and MAC Header

The IEEE 802.11a [6] is the amended specification to IEEE 802.11. This specificationdescribes a high speed physical layer (PHY) which operates in the 5 GHz band. Thetechnique used in the 802.11a is the orthogonal frequency division multiplexing (OFDM)with a rate capabilities of 6, 9, 10, 12, 18, 24, 34, 36, 48 and 54 Mbits/s. The PHYheader used in the PHY layer is called Physical Packet Data Unit (PPDU) as shown inthe fig 6.4. PHY layer receives frame from the MAC layer and add more informationbefore transmission. PHY header consists of three parts; preamble, signal and Data asshown in the fig 6.4. The preamble is used for the synchronization. Signal field is usedto determine the frame length. The data field consists of frame which is received fromthe MAC layer.MAC frame consists of a MAC header and cyclic redundancy check (CRC) and data.

Figure 6.4: Physical data packet

6.5 EMCD Behavior in Different Scenarios

In this section, EMCD behavior is analyzed in different scenarios. First, the study aboutEMCD, how it works and detects the collisions among multiple multicasts, multicast andunicast senders. Secondly, discussion about those cases in which collision is not detected.

6.5.1 Collision between two Multicast Senders

Fig 6.5 shows two multicast senders that start their transmission at the same time andexplains how the collision is detected between them. AP1 detects the collision beforeAP2 due to its smaller Tv. AP1 starts its carrier sensing phase before AP2 and findsthe medium busy and jams the medium. If AP1 does not jam the medium then AP2starts its carrier sensing phase and finds the medium idle. So jamming the medium,ensures that the potential transmitters also know about collision occurrence. At theend of TCDI , normal DCF procedure is invoked and retransmission is scheduled. CWis not doubled here due to different factors. When a station finds the medium idle, thenormal transmission will be started.

6.5.2 Collision between Unicast and Multicast Senders

Fig 6.6 shows that simultaneous transmission from a unicast station and a multicaststation leads to a collision. The AP senses the station and comes to know that the

28 Chapter 6. Early Multicast Collision Detection

Figure 6.5: Collision detection between two multicast senders

medium is busy and it transmits the collided packet in order to jam the transmission.So the multicast sender detects the collision due to its EMCD collision sensing be-havior, but unicast station detects collision by following the normal DCF mechanism.Retransmission is also scheduled for both multicast and unicast senders by normal DCFmechanism.

Figure 6.6: Collision detection between unicast and multicast senders

6.5.3 Undetectable Collisions

There are two possibilities in which EMCD can not detect collisions. First, if theunicast packet size is smaller than the Tv size as shown in fig 6.7. Secondly, if twomulticast senders have the same size of the Tv and try to transmit them simultaneously,as illustrated in the fig 6.8.

Figure 6.7: Undetectable collision between unicast and multicast senders

6.5. EMCD Behavior in Different Scenarios 29

Figure 6.8: Undetectable collision between two multicast senders

30 Chapter 6. Early Multicast Collision Detection

Chapter 7

Design and Implementation

7.1 Introduction to GloMoSim

Global Mobile Information System Simulator (GloMoSim) [26] is a scalable simulationenvironment designed for large mobile and wireless communication networks, developedat The University of California, Los Angeles (UCLA) Parallel Computing Laboratory.GloMoSim is a discrete-event simulator which is capable of performing parallel simula-tion provided by the PARSEC (Parallel Simulation Environment for Complex Systems)[10]. PARSEC is a C based simulation language, also developed by the UCLA ParallelComputing Laboratory, designed for sequential and parallel execution of discrete-eventsimulation models [17].

Layers Protocols

Mobility Random waypoint, Random drunken, Trace based

Radio Propagation Two ray and Free space

Radio Model Noise Accumulating

Packet Reception Models SNR bounded, BER based with BPSK/QPSK

modulation

Data Link (MAC) CSMA, IEEE 802.11 and MACA

Network (Routing) IP with AODV, Bellman-Ford, DSR, Fisheye,

LAR scheme 1, ODMRP, WRP

Transport TCP and UDP

Application CBR, FTP, HTTP and Telnet

Table 7.1: Protocols and Models supported by GloMoSim at each layer

GloMoSim can be used to simulate networks with up to thousand mobile nodesconnected by a heterogeneous wireless networks that includes multicast, multi-hop adhoc networks, asymmetric communication using direct satellite broadcast and traditionalInternet protocols. GloMoSim is built by using a layered approach which is similar tothe OSI seven layer network architecture and standard APIs are used in the simulationlayers. That makes it easy to implement new protocols and models at different layers.The protocols supported by GloMoSim used by the different simulation layers are shownin table 7.1.

GloMoSim is a discrete event driven simulator that means the execution of the

32 Chapter 7. Design and Implementation

simulation model consists of a set of events. The events occur and an appropriateaction has been taken in its response. The event is defined as an incident which makesthe root to change the state of the system. A certain event or a group of events maycause other events to be invoked and so on and that is how the simulation proceeds.Events can only occur at different time units and it is not allowed to occur betweenthese time units. An event can be an arrival of packet at some layer or expires sometime period.

GloMoSim is freely available for educational purposes, so-called academic versionand can be accessed from an educational domain. The academic version only supportsthe sequential execution of simulations.

7.2 Design and Implementation

In this section, an overview of design and implementation of EMCD and PREMA ispresented. GloMosim supports IEEE 802.11 together with many other MAC protocols,but none of them is achieving successful results for multicast. Thus, the task was to de-sign and implement the two reliable multicast schemes. The following section illustratesthe important components used for the implementation of EMCD and PREMA.

7.2.1 Implementation of EMCD

EMCD is implemented at the MAC layer of GloMoSim to provide multicast functionality.GloMoSim is built by using the layered approach. It is really easy to understand and ismore flexible to use than other common simulators like NS2 etc. Data Display Debugger(DDD v3.11) helped to understand the GloMoSim structure by follow the sequenceof instructions at the different layers. The common notations which are used for theimplementation of EMCD i.e D1 and D2 represents the Tv and Tm respectively. Thefollowing sections illustrates the important components used for the implementation ofEMCD.

GlomoMacEmcd

GlomoMacEmcd is a data structure that holds important data members and representsthe MAC of the entire station. GlomoMacEmcd uses the GlomoMac802 11 with somemodifications as listed in the table 7.2.

GlomoMacEmcd data structureclocktype CW FOR MULTICASTNODE ADDR destAddr TempBOOL isChannelIdleBOOL isD1Transmittedfloat D1sizefloat Txminint payLoadSizeint packetSizeint collisions

Table 7.2: Important data members used in GlomoMacEmcd data structure

7.2. Design and Implementation 33

CWFOR MULTICAST holds information about the CW which is used for the mul-ticast stations. destAddr Temp is an address parameter which describes about thetransmission (unicast or multicast). The two flags isChannelIdle and isD1Transmittedare used to kept the information about channel (busy or idle) and Tv. The decision ofthe Tm is based on the flag isD1Transmitted, if the flag is TRUE then Tm will be trans-mitted otherwise not. The four parameters D1size, Txmin, payLoadSize, packetSizeand collisions are used to kept the information of the size of Tv, transmission minimum,packet size of Tv and payload size respectively. The collisions variable is used to containthe information of collisions detected on the MAC layer by the EMCD.

Emcd MacFrameType

Emcd MacFrameType uses 802 11MacFrameType with enhancements as listed in table7.3. There are three MAC frame types used in the Emcd MacFrameType data structure.EMCD D1, EMCD D2, EMCD COLLIDE are the frames types which represents Tv, Tmand collided transmissions.

Frame typesEMCD D1EMCD D2EMCD COLLIDE

Table 7.3: Frame types used in Emcd MacFrameType data structure

EMCD MacStates

In EMCD, different states are defined in which a station can be switched. These statesare transmitting and waiting states. More states are added to the original standard ofthe GloMoSim as listed in table 7.4.Tv is transmitted by setting the state to EMCD X D1. After Tv, control comes to thesecond phase of listening by changing state to EMCD S WFIFS. After CIFS time pe-riod, state is changed with EMCD S WFNEXTIFS and TransmitDataFrame() methodis called. Now, if the channel is sensed idle then Tm will be done otherwise col-lided transmission will be happened by changing their states to EMCD X D2 andEMCD D2PktCollided.

Different StatesEMCD S WFIFSEMCD S WFNEXTIFSEMCD X D1EMCD X D2EMCD X D2PktCollided

Table 7.4: Important states used in Emcd MacStates data structure

34 Chapter 7. Design and Implementation

7.2.2 Working of EMCD

The design is based on the EMCD functionality. In this algorithm, a transmission isdivided into two parts, first transmission takes place and a pause follows after a fixedtime interval. Secondly, the channel is sensed for a while, if the channel is idle then thetransmission is done.As discussed earlier that solution is based on three major phases. The work is toimplement EMCD algorithm at the MAC layer of GloMoSim. The functionality of theMAC layer starts from the point when a packet arrives from the network queue to theMAC layer. The control moves forward by setting the states and enters into the backoffmechanism if network queue is not empty. If the queue is found empty then the controldirectly goes for the transmission of the packet after changing its state.

Figure 7.1: Sequence diagram of EMCD with successful transmission

7.2. Design and Implementation 35

Vanguard Transmission

The role of the EMCD is involved from the point where traffic is decided to be a unicastor multicast. The function of the EMCD algorithm starts from the point when traffic isdetected to multicast in the function MacEmcdTransmitDataFrame() as shown in thefig 7.1. The packet is taken from the queue and a new packet is created according theequation 6.1 and transmitted.

Figure 7.2: Sequence diagram of EMCD with unsuccessful transmission

36 Chapter 7. Design and Implementation

Pause and Channel Sensing

When the control arrives at the function TransmissionHasfinished(), the next phaseof sensing starts. In the function MacEmcd D1PKT Transmitted() which is called byTransmissionHasfinished(), WFIFS state setting and timer initialization is done by call-ing MacEmcdSetState() and EmcdStartTimer() functions respectively. The CIFS valueused in EmcdStartTimer() is fixed which is 10µS to carry out a proper CCA operation.The channel is sensed twice here before and after calling of EmcdStartTimer() function.

Collided/Main Transmission

The start of third phase is the point when the control arrives at the function MacEm-cdTransmitDataFrame() after channel sensing twice. If the channel is sensed idle thencontrol goes to the second section of the function of TransmitDataFrame(), generatesthe D2 packet which is known as Tm and transmits that as shown in the last section offig 7.1. But if the channel is reported to be busy then D2 collided packet is generatedand is transmitted. The last section of fig 7.2 is showing the sequence to deal with D2collided packet.

7.2.3 Implementation of PREMA

The PREMA is implemented on the MAC layer in GloMoSim v2.03. IEEE 802.11 andsome other MAC protocols are currently supported by GloMoSim. Our task was todesign and implement PREMA on the existing DCF mechanism. The following sectiondescribes the important components that are used for the implementation of PREMA.

GlomoMacPrema

The GlomoMacPrema is modified version of GlomoMac802 11 data structure by addingthe features of PREMA. The important parameters of GlomoMacPrema is illustratedin table 7.5.

PREMA Data Structureint stateint prevStateclocktype CWclocktype BOint idleSlotBOOL isAFirstBurstlong BurstLength

Table 7.5: Important parameters of GlomoMacPrema

idleSlot kejpg the information of number of eliminations that how many eliminationsof particular station have been done. isAFirstBurst is used as tag to perform certaintasks, like reseting the values etc. As name indicates the BurstLength is used to keptthe length of burst which is sampled from the geometric distribution.

Some constant parameters are also defined, as shown in table 7.6 with default val-ues. PREMA THRESHOLD is a total number of elimination phases and PI is used ingeometric distribution.

7.2. Design and Implementation 37

PREMA ParametersPREMA THRESHOLD 4PREMA PI 0.5

Table 7.6: Some constant parameters used in implementation of PREMA

PREMA MacFrameType

There are many frame types defined at the MAC layer. Some modified frame types forPREMA are describes in table 7.7.

PREMA Frame TypesPREMA BURSTPREMA BROADCAST

Table 7.7: Frame types used in implementation of PREMA

Prema MacStates

There are several states are defined in GloMoSim. The MAC layer states are dividedinto three categories, wait for response states, transmission states and miscellaneousstates. There are three states added for PREMA in GloMoSim, as illustrated in table7.8.

PREMA StatesPREMA X BURSTPREMA X MULTICASTPREMA X WFCS

Table 7.8: MAC layer states used in implementation of PREMA

PREMA X BURST and PREMA X MULTICAST are the transmission states. Thestation enters into these states when it going to transmit a burst or multicast the frame.PREMA S WFCS indicates the time period for carrier sensing comes between two elim-ination phases.

7.2.4 Working of PREMA

The working of PREMA is divided into three phases, transmission of the burst, carriersensing and transmission of the actual packet. The sequence of PREMA with functionsand some important parameters, is shown in fig 7.3.

When the frame arrives at the MAC layer, the station attempts to go to the DIFSby performing the carrier sensing. If the medium is idle, the state of the system changesand the DIFS time period starts. If there is only one frame in the queue then the backoffwill be zero and directly goes for the transmission. Otherwise the backoff timer startsand transmits packet when backoff reaches to zero. At this point, PREMA operationstarts.

38 Chapter 7. Design and Implementation

Figure 7.3: Sequence diagram of PREMA

Burst Transmission

After the backoff time period, the station enters into the bursting phase as shown in fig7.3. The PREMA gets the frame from the network layer. Only multicast stations enterinto the bursting phase. The multicast stations sample the burst length by calling theGeometric Distribution function. After changing to the state ’PREMA X ’ BURST, thetransmission of burst starts with the sampled length.

Carrier Sensing

The radio layer informs the change of status to the MAC layer. The MAC layer callsPREMA to take appropriate actions. The radio status informs that the transmissionof the burst has finished. The next step of PREMA is to perform the carrier sensingafter waiting a very short time period. It is because the minimum signal detection

7.2. Design and Implementation 39

time is approximately 4µs and it is useless to perform carrier sensing immediately afterthe transmission. The state is changing to PREMA S WFCS. As the time expires thestation performs the carrier sensing by verifying the radio status. If the medium isidle the station performs another bursting and then carrier sensing until the specifiedrepeated eliminations.

Frame Transmission

After the repeated eliminations, the winner starts the transmission of the actual frame.Again PREMA calls the network layer but this time the frame is dequeued. Since theretransmission is not possible for multicast frame so there is no meaning to be still inqueue. The state changes to PREMA X BROADCAST and the transmission of thedequeued frame starts.

40 Chapter 7. Design and Implementation

Chapter 8

Evaluation

8.1 Introduction

In this chapter, the performance evaluation of EMCD and PREMA is analyzed and con-clusions are made on the results of the simulated scenarios. The evaluation presentationis divided into three parts, 1) EMCD performance evaluation, 2) PREMA performanceevaluation and 3)comparison between them.In first section, the EMCD is simulated with some simple scenarios to observe the roleof affecting parameters to ensure the reliability of multicast transmissions. In the sec-ond section, PREMA is simulated and analyzed with simple scenarios and highlights itsperformance affecting parameters for the multicast transmission. In the third section,the comparative evaluation of EMCD, PREMA and 802.11 is presented by adapting amore realistic approach in the simulations. Last section describes who is more suitablefor multicast traffic in real time scenarios.

8.2 Performance Parameters

The performance evaluation is presented in the following QoS metrics.

– Throughput Throughput is the rate of data transferred over the time and is ex-pressed in bits per second. It is really difficult to show the multicast through-put precisely, because in multiple receivers, few of them may not receive packets.Therefore in these simulations, the packet received by one receiver in a multicastgroup is considered to be received by all receivers. Hence, aggregated throughputis taken for the result presentation.

– Collision A collision is happened, when two or more senders try to transmit data atthe same time. Collisions result in retransmissions which also lead to the wastageof bandwidth and decreases the throughput.

– Channel Utilization It is an achieved throughput presented in percentage, is lessambiguous term for the presentation of achieved throughput.

42 Chapter 8. Evaluation

8.3 Simulation Model

8.3.1 Network Architecture

In simulations, the Basic Service Set (BSS) infrastructure has been used which consistsof different number of APs and stations. Hence, all the communication takes placethrough the AP, therefore, the stations are not able to communicate directly.

8.3.2 Physical Parameters

IEEE 802.11a has been used for simulations which is based on the Orthogonal Frequency-Division Multiplexing (OFDM) specification. There are some parameters that are usedfor OFDM specification as listed in table 8.1.

Parameter Value

T4 4µs

TCDI 112µs

CIFS 10µs

DIFS 34µs

SIFS 16µs

CWmin 15

CWmax 1023

PLCP Preamble 16µs

Signal 4µs

ACK 24µs

Distance between Access Points 50 m

Transmit Power 40 mW

Simulation Time 50µs

Rate used for multicast 6Mbps

Rate used for unicast 6Mbps

CCA Time 4µs

Table 8.1: EMCD parameters used in simulations

8.4 Evaluation of EMCD

8.4.1 Throughput

In this section the performance of EMCD is measured in terms of throughput. EMCDis an early collision detection method which pauses transmission and listen for channel.Therefore the performance of EMCD is highly dependent on the duration of the TCDIas discussed in section 6.3. Another important factor is backoff mechanism which isused for retransmission and also affects the system throughput.

Collision Detection Interval

Different TCDI values and increase in stations from (2-10) have been used in this sim-ulation. The ratio between TCDI and number of selectable are listed in the table 8.2.

8.4. Evaluation of EMCD 43

The CW is fixed to 15. The rest of values are used in the simulation according to thetable 8.1.

Collision Detection Interval TCDI No. of Selectable

72 0

80 2

112 10

152 20

200 30

Table 8.2: EMCD TCDI values used in simulation

Figure 8.1: EMCD throughput on different TCDI values.

Discussion on ResultThe fig 8.1 shows the results and further illustrates that how much the TCDI is

important in the performance of EMCD. The simulation result is described on the fol-lowing three points.

– Smaller TCDI Value

Smaller TCDI values lead to small number of choices for unique transmissions.At the TCDI 72, the choice for unique transmission is zero and the packet withphysical and MAC header is transmitted, which leads to the undetectable colli-sions, that is illustrated in fig 6.8. As there is no successful transmission because

44 Chapter 8. Evaluation

of collisions, so the throughput is zero. At the TCDI 80, the choice for uniquetransmission is 3. Since the detection ratio is increased due to more choices forunique transmissions as shown in fig 8.1. Due to the increase in detection ratio,the throughput is also increased.

– Higher TCDI Value

Higher TCDI value leads to more choices for the unique transmissions but it alsocauses increasing in the size of that unique transmission. When the collisionoccurs for that longer unique transmission, it causes longer delay and decrease inthroughput. Fig 8.1 shows the decrease in throughput with increase of selectablese.g., 20 and 30.

– Optimal TCDI Value

As stations between 2 to 10 are utilizing the best channel resources on the selectablevalue 10 with TCDIvalue 112. Fig 8.1 result can be summarized as there is a tradeoff relation between the number of stations and the number of selectable. In otherwords, throughput is proportional to the collision detection ability.

Contention Window (CW) value

Optimal CW is one of the major factor that also increases the channel utilization. Italways depends on the CW value. In this algorithm, the focus is about the effect of theretransmission on throughput but not about different contention schemes.

Same values has been used in the simulation, which are listed in table 8.2. For thissimulation, the number of stations increases from 10 to 100, and the CW values are 15,31, 63, 100. The value for the TCDI is 112.

Figure 8.2: EMCD throughput on different CW values.

8.4. Evaluation of EMCD 45

Discussion on ResultsFig 8.2 shows surprising results, the higher CW values (63,100) show the high

throughput when the number of stations are few and constantly decreasing with theincrease of number of stations. On the other hand, smaller CW values (15,31) achievelow throughput in the start and later maintains its throughput. If CW values are higherthen the probability of collision is very low which improves performance of EMCD. Butwith the increase in the number of stations, the probability of collision is also increased.So at the higher CW value, the contention phase is relatively large that leads to theprobability of longer delays and in smaller values of CW, the case is inverse. Inconsis-tent results in this figure show the aggregated affect of both TCDIand CW value on thethroughput.

8.4.2 Collisions

In this section we study the comparison between detection and undetection collisionprobability at different number of selectables. The results are taken in Matlab by usingfollowing formulas.

Undetectable

UndetectionProbability = (1/N)k−1 (8.1)

Figure 8.3: Undetection probability at different number of selectables

46 Chapter 8. Evaluation

Detectable

DetectionProbability = 1− (1/N)k−1 (8.2)

Figure 8.4: Detection probability at different number of selectables

Here N represents the number of stations and k represents the number of slots avail-able or selectables.

Discussion on ResultsThe results are presented in fig 8.3 and fig 8.4. The ideal conditions for simulation is

assumed. Queue is full and is always ready to send the packets. Only multicast stationsare assumed. The collision detection capability increases with the increasing number ofstations and the number of selectables. The detection capability is zero when there isno choice for the unique transmission, and the undetection is the inverse of detectioncapability.

TCDI Collisions

The simulation for collision detection on the MAC layer is the same simulation havingthe same parameters used for the throughput in the fig 8.1. The Figure is clearlyshowing the relationship between detection ratio and throughput. As we discussed insection 8.4.1, throughput increases with the increase in detection ability. The detectionability is high at the optimal TCDI value 112 as shown in the fig 8.5.

8.5. Evaluation of PREMA 47

Figure 8.5: Collision at MAC layer for vanguard transmissions

8.5 Evaluation of PREMA

8.5.1 Effecting Parameter

This section presents the performance evaluation of PREMA on the bases of through-put. The evaluation of PREMA enables us to observe the effectiveness of parametersh and q, as described in section 5.1.3, and helps to find out the optimal value for theseparameters.A simple scenario is considered in order to get optimal parameters for PREMA. In thissimulation the number of stations is fixed to 50. The value of parameter q is increasingfrom .1 to .9 during the simulation. Therefore, eight different values of h are tested foreach corresponding values of q. Similar results are presented in [12].

Discussion on ResultAs Gerger et al. suggests [12] and fig 8.6 shows that the optimal values of h and q

is 4 and 0.5, respectively. The throughput is at maximum at these points. Lower valuesof q gives lower throughput. By using the lower value of q, the burst length is largeand the throughput is wasted. As we show in the graph, the throughput is increasedby the increase of q up to certain point 0.8 and then decreases. On the other hand, forthe lower values of h means that less number of elimination phases, can not resolves theconflicts completely and throughput is low. By the increase of elimination phases,

48 Chapter 8. Evaluation

Figure 8.6: Prema results on different values of q and h parameters

throughput increases. From the 4th line of h as shown in fig 8.6 , there is no significantincrease in throughput. By using four elimination phases, there is large probability of asingle winner, so that the remaining elimination phases are wasted. In general, for theoptimal lower value of h corresponds to the higher value of q and vice versa.

8.6 Comparison

In this section, the comparison of EMCD, PREMA and 802.11 is presented. The aimof this study is to distinguish the algorithms and to focus, which one is more reliableand efficient for the multicast traffic. The results are presented into two sections. In thefirst section, the throughput comparison for multicast traffic is presented. In the secondsection, more realistic scenarios are used with fixed number of broadcast stations andvarying number of unicast stations.

8.6.1 Throughput Comparison of Multicast Traffic

The simulation is done with increasing number of stations from (10-100). The optimalvalues are used for this simulation to make the comparison. The EMCD TCDIoptimalvalue 112 is used for this simulation. The optimal values used for PREMA are h=4 andq=0.5. The rest of values which are used are mentioned in table 8.2.

Discussion on ResultsThe PREMA outperforms the 802.11 and EMCD in all scenarios as shown in the

fig 8.8 . The repeated elimination phases give the high throughput. The probability ofa unique winner for the transmission after the eliminations is approximately 99%. ThePREMA sustains the throughput during the simulation, and has effect of increase in thenumber of stations. That shows clearly the PREMA is the best option for the multicasttraffic to any scale of network.

8.6. Comparison 49

Figure 8.7: Throughput comparison of PREMA, EMCD and 802.11

8.6.2 Throughput Comparison among Multicast and UnicastStations

There are 10 multicast stations and increasing number of unicast station from 10 till 50,used in this simulation. The simulation starts from 10 multicast stations and 10 unicaststations are added simultaneously. The other values used in the simulation are listed inthe table 8.1. last simulation result. The results are presented in fig 8.8. Both are partof the same simulation scenario.

Figure 8.8: Throughput comparison of PREMA, EMCD and 802.11.

Discussion on ResultsAs shown in the fig 8.8, fairness among the unicast and multicast traffic is relatively

higher for PREMA. The results show that EMCD multicast stations are utilizing morechannel resources as compared to other algorithms. The reason for this behavior thatEMCD does not double CW value on collisions. Therefore, EMCD wins more channelaccess.

50 Chapter 8. Evaluation

Chapter 9

Conclusion and FurtherResearch

In this thesis, We have studied several existing problems in order to use the multicasttraffic in IEEE 802.11 networks. The lack of the ARQ protocol, unfairness amongunicast and multicast traffic, and multiple multicast traffic in the same ESS causes highcollision rate and poor performance. In this regard, IEEE 802.11 is unsuitable optionfor the multicast traffic in wireless networks.EMCD and PREMA are the focus of our study from different approaches for dealingwith the multicast problems. The simulation results present the key parameters toimprove the performance of both algorithms.The comparison between EMCD, PREMAand 802.11 algorithm is also presented.

9.1 EMCD

EMCD minimizes the overall packet loss rate and enhances throughput. EMCD per-formance is highly dependent on the TCDI value and at some extent on the CW value.The results are concluded and summarized as.

– As shown in the results, the EMCD performance has a direct proportion to itsdetection capability. Detection ability is highly dependent on the throshlod value.

– TCDI has a trade-off relation with the network. So, it always demands investiga-tion of optimal TCDI value according to the scale of network.

– EMCD algorithm has combined effect of two phases to solve the collision, Phase-Ibackoff phase and Phase-II EMCD collision detection. So its efficiency is dependenton the TCDI and CW value.

9.1.1 Advantages

– EMCD achieves high channel utilization in small scale networks.

– EMCD based on collision detection on feedback which may minimize the hiddenterminal problem in the multicast traffic[24].

52 Chapter 9. Conclusion and Further Research

9.1.2 Disadvantages

– EMCD does not work very well with the increase of the network load.

– EMCD introduces additional overhead.

– EMCD is expensive and complex to perform CCA operation during transmission.EMCD is harder to implement due to a pause during transmission, sense thechannel and then again transmission.

9.2 PREMA

PREMA is a collision resolution algorithm that resolves collisions by jamming the chan-nel with brusts. Two parameters h and q are used for the elimination phases and theprobability q is used in the geometric distribution to sample the burst length, respec-tively. PREMA resolves the collision between multicast stations as well as betweenunicast and multicast. The conclusions of the results are summarized as.

9.2.1 Advantages

– PREMA achieves very high channel utilization and success probability.

– The nature of PREMA is memoryless, i.e. each elimination phase is independent,that provides good fairness.

– Burst length and the number of elimination phases are adjustable according to thescenario to achieve the maximum channel utilization.

– PREMA performs well in large scale networks.

9.2.2 Disadvantages

– PREMA is not a good option to deal with the hidden terminal problem

– PREMA is expensive too in terms of power loss during its jamming session.

9.3 Further Reading and Research

There are several issues with the proposed algorithms, that needs to be further investi-gated. Some of them are listed below

9.3.1 EMCD

– EMCD in terms of bursting efficiency [16].

– EMCD starts with contention phase and then moves to the collision detectionphase to accomplish its task. Therefore it is dependent on both values. To improvethe efficiency in EMCD, further study is needed to do research on alternativesolutions about combining both phases (contention + collision detection).

9.3. Further Reading and Research 53

9.3.2 PREMA

– PREMA outperforms EMCD and IEEE 802.11 in throughput but is expensive inpower utilization. SEMA is the motivated study for this issue [16].

54 Chapter 9. Conclusion and Further Research

Chapter 10

Summary

In this thesis, our aim was to study the multicast services in IEEE 802.11 networks.IEEE 802.11 is a popular technology in the wireless communication world due to its lowcost and ease of deployment. Mostly DCF basic access mechanism is used for the IEEE802.11 networks, which is based on CSMA/CA ”listen before transmit” scheme.In this study, problems associated with multicast in IEEE 802.11 networks have beeninvestigated. The lack of ARQ protocol, unfairness among unicast and multicast trafficand multiple multicast traffic in the same ESS leads to high collisions and poor perfor-mance.Different approaches have been studied to overcome the problems in multicast trans-missions. There are a diversity of algorithms to solve the problems which are organizedunder the collision resultion, collision detection and channel reservation approaches.The focus of this study is on the two algorithms EMCD and PREMA using collisiondetection and collision resultion approaches respectively.EMCD is the collision detection method which performs listening operation during trans-mission to send a packet. EMCD is composed of three phases: 1.) Tv 2.) Carrier sensing3.) main/collided transmission. The TCDI value is the key parameter to attain highperformance. Correct values of CW also enhances performance.PREMA is a simple collision resolution algorithm that uses bursting and repeated elimi-nation to resolve the channel conflicts for multicast transmission. Burst is used for chan-nel jamming, and elimination phase determines the surviving station with the longestburst. PREMA is modified and integrated with IEEE 802.11, so that multicast trans-mission is handled by PREMA and unicast by IEEE 802.11. There are two optimalparameters of PREMA, h and q, where h is the number of elimination phases and q isthe probability used in the geometric distribution to sample the burst length. PREMAresolves the collisions between multicast stations as well as between unicast and multi-cast.The results presented in this study, show that EMCD and PREMA improves the per-formance of multicast. The comparative study shows that PREMA outperforms EMCDand IEEE 802.11 in all scenarios due to its independence relevant to the scale of thenetwork.

56 Chapter 10. Summary

Appendix A

List of Abbreviations

AP Access PointAIFS Arbitration Inter Frame SpaceAODV Ad-hoc On-demand Distance VectorAP Access PointAPI Application Programming InterfaceARQ Automatic Repeat RequestBEB Binary Exponential BackoffBER Bit Error RateBO Back offBPSK Binary Phase Shit-KeyingBSS Basic Service SetBLBP Beacon Driven Leader Based ProtocolCA Collision AvoidanceCBR Constant Bit RateCCA Clear Channel AssesmentCIFS Collision Inter Frame SpaceCSMA/CA Carrier Sense Multiple Access / Collision AvoidanceCW Contention WindowCWmax Contention Window MaximumCWmin Contention Window MinimumDCF Distributed Coordination FunctionDDD Data Display Debugger

58 Chapter A. List of Abbreviations

DIFS DCF Inter Frame SpaceDS Distribution SystemDSSS Direct Sequence Spread SpectrumEMCD Early Multicast Collision DetectionESS Extended Service SetEY-NPMA Elimination Yield Non-Preemptive Prioritized Multiple AccessFTP File Transfer ProtocolFHSS Frequency Hopping Spread SpectrumGloMoSim Global Mobile Information System SimulatorHiperLan High Performance Radio LanHCF Hybrid Coordination FunctionHTTP Hypertext Transfer ProtocolIBSS Independent Basic Service SetIP Internet ProtocolISM Industrial Scientific and MedicalIR Infra RedIFS Inter Frame SpaceMAC Medium Access ControlMbps Mega bit per secondMC Multicast StationODMRP On-Demand Multicast Routing ProtocolOFDM Orthogonal frequency-division multiplexingPARSEC Parallel Simulation Envoirnment for Complex SystemsPCF Point Coordination FunctionPIFS Point Inter Frame SpacePREMA Prioritized Repeated Elimination Multiple AccessQoS Quality of ServiceSEMA Silent Elimination Multiple AccessSIFS Short Inter Frame SpaceSLBP Simple Leader Base ProtocolSNR Signal to Noise RationSTA StationTCDI Collision Detection IntervalTCP Transmission Control ProtocolTIFS Time Medium- Inter Frame SpaceTxmin Transmission MinimumUC Unicast StationUCLA University of California, Los AngelesUDP User Datagram ProtocolWFIFS Wait for Inter Frame SpaceWifi Wireless FidelityWLAN Wireless Local Area NetworkWRP Wireless Routing Protocol

References

[1] IEEE 802.11. http://en.wikipedia.org/wiki/IEEE 802.11, accessed 2008-05-15.

[2] Multicast MAC extensions for high rate real-time traffic in Wireless LANs.http://www.nt.uni-saarland.de/projects/mac/, accessed 2008-02-15.

[3] IEEE Std. 802.3, Part 3: Carrier Sense Multiple Access with Collision Detection(CSMA/CD) Access Method and Physical Layer Specifications. 1985.

[4] IEEE Std. 802.11, Part 11: Wireless LAN Medium Access Control (MAC) andPhysical Layer (PHY) Specifications. 1997.

[5] Broadband Radio Access Networks (BRAN); High Performance Radio Local AreaNetwork (HiperLAN) type 1; functional specification. 1.2.1(ESTI), 1998.

[6] IEEE Std. 802.11a, Supplement to Part 11: Wireless LAN Medium Access Con-trol (MAC) and Physical Layer (PHY) Specifications: Higher-Speed Physical LayerExtension in the 5 GHz Band. 1999.

[7] IEEE Std. 802.11b, Supplement to Part 11: Wireless LAN Medium Access Con-trol (MAC) and Physical Layer (PHY) Specifications: Higher-Speed Physical LayerExtension in the 2.4 GHz Band. 1999.

[8] Wireless LAN Medium Access Control (MAC) and Physical Layer Specifications,1999. www.csse.uwa.edu.au/adhocnets/802.11-1999.pdf , accessed 2007-05-12.

[9] I. Aad, C. Castelluccia, and R.A. INRIA. Differentiation mechanisms for IEEE802.11. INFOCOM 2001. Twentieth Annual Joint Conference of the IEEE Com-puter and Communications Societies. Proceedings. IEEE, 2001.

[10] Rajive Bagrodia, Richard Meyer, Mineo Takai, Yu an Chen, Xiang Zeng, Jay Mar-tin, and Ha Yoon Song. PARSEC: A Parallel Simulation Environment for ComplexSystems. IEEE Computer, 31(10):77–85, October 1998.

[11] Jahanzeb Farooq and Bilal Rauf. Implementation and Evaluation of IEEE 802.11eWireless LAN in GloMoSim. pages 1–8, Department of Computing Science, UmeaUniversity, Umea, Sweden, 2006.

[12] Thomas Nilsson Greger Wikstrand. Prioritized Repeated Elimination MultipleAccess: A Novel Protocol for Wireless Networks. Conference of the IEEE ComputerCommunications., 27(INFOCOM.), 2008.

59

60 REFERENCES

[13] S.K Kasera J. Kuri. Reliable multicast in multi-access wireless LANs. EighteenthAnnual Joint Conference of the IEEE Computer and Communications Societies.Proceedings. IEEE, 7(INFOCOM ’99.):760–767, 1999.

[14] Jean Tourrilhes. Robust Broadcast: Improving the reliability of broadcast trans-missions on CSMA/CA. Hewlett Packard Laboratories, page 5, 2002.

[15] J. Kuri and S.K. Kasera. Reliable Multicast in Multi-Access Wireless LANs. Wire-less Networks, 7(4):359–369, 2001.

[16] Thomas Nilsson. Distributed Multiple Access and Service Differentiation Algorithmsfor Wireless Networks. Umea University, SE-901 87 Umea Sweden, PHD Thesis,Dept. of Computing Science, Umea University, January 2008.

[17] J. Nuevo. A Comprehensible GloMoSim Tutorial, 2003.

[18] Raphael Rom. Local area and multiple access networks. Computer Science Press,Inc., New York, NY, USA, 1986.

[19] Jochen Schiller. Mobile Communications. Addison-Wesley, second edition, 2003.

[20] J.L. Sobrinho and AS Krishnakumar. Medium Access Control Layer. Bell LabsTechnical Journal, 10:173, 1996.

[21] Gerla .M Tang .K. MAC reliable broadcast in ad hoc networks. Military Commu-nications Conference, 2001. MILCOM 2001. Communications for Network-CentricOperations: Creating the Information Force. IEEE, 2(2):1008– 1013, 2001.

[22] Jerry Eriksson Thomas Nilsson, Greger Wikstrand. Collision detection method formulticast transmissions in CSMA/CA networks. Wireless Communications andMobile Computing, 7(6):795 – 808, 7 Jul 2006.

[23] Lennart Bondesson Thomas Nilsson, Greger Wikstrand. Silent Elimination MultipleAccess: An Efficient Channel Brusting Protocol.

[24] Greger Wikstrand. Human Factors and Wireless Network Applications More Bitsand Better Bits. Technical Report UMINF 06.34, Dept. of Computing Science,Umea University, Umea, Sweden, 2006.

[25] H. Wing Lo, Mouftah. Collision Detection and Multitone Tree Search for Multiple-Access Protocols on Radio Channels. Selectd Areas in Communications, 5(6):1035– 1040, Jul 1987.

[26] X.Zeng, R. Bagrodia, and M. Gerla. GloMoSim: A Library for the Parallel Simu-lation of Large scale Wireless Networks. In Proceedings of the 12th Workshop onParallel and Distribution Simulation PADS, 1998.