<%BANNER%>

Performance Enhancement of CSMA/CA in Powerline Communications under Heavy Traffic and Hidden Node Conditions

Permanent Link: http://ufdc.ufl.edu/UFE0022857/00001

Material Information

Title: Performance Enhancement of CSMA/CA in Powerline Communications under Heavy Traffic and Hidden Node Conditions
Physical Description: 1 online resource (110 p.)
Language: english
Creator: Lee, Jongdae
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: Electrical and Computer Engineering -- Dissertations, Academic -- UF
Genre: Electrical and Computer Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: A communication protocol is a set of standard rules for data representation, signaling, authentication, and error detection required to send information over a communication channel. Historically, we have used various media as communication channels; copper wire, coaxial cable, air, power lines, and various protocols have been standardized for using those channels based on the characteristics of communication. HomePlug 1.0 protocol, the first and most commonly used technology for power line communication, is based on a Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA), where a station listens to the medium before transmission to avoid collision. Due to the contention property of CSMA/CA protocol, the performance of HomePlug 1.0 MAC becomes critical when the number of users increases, just as in 802.11 wireless MAC protocol. Another big issue in a CSMA/CA-controlled network is a hidden terminal problem which occurs because the carrier sensing range is different from the transmission range. This hidden terminal problem increases the potential collision probability which dramatically degrades the network performance. Based on the original HomePlug 1.0, we proposed an Optimal Constant Contention Window (OCCW) mechanism which dynamically selects the best fixed contention window size according to the number of contending stations in the channel. An analytical and simulation framework was used to find the optimal values of contention window size for best performance, and it is shown that using the proposed scheme results in almost constant throughput of about 80% independent of the number of nodes. Also we implemented a node estimation algorithm that each station can use independently to find the number of contending stations. The simulation results demonstrated that this estimation algorithm was well tracking the predefined actual number of nodes; moreover, it provided much better performance than the original HomePlug 1.0 protocol even under some hidden terminal conditions. Another part of research, an efficient TCP mechanism by reducing out-of-sequence packets which may happen during handover procedure in mobile IPv6 has been proposed at the end of this dissertation. A packet reordering algorithm based on the snoop protocol has been proposed to arrange the out-of-sequence packets and simulation results demonstrated that managing the packet sequence increases the overall TCP performance by reducing the packet re-transmission in Mobile IPv6 network.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Jongdae Lee.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.
Local: Adviser: Latchman, Haniph A.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022857:00001

Permanent Link: http://ufdc.ufl.edu/UFE0022857/00001

Material Information

Title: Performance Enhancement of CSMA/CA in Powerline Communications under Heavy Traffic and Hidden Node Conditions
Physical Description: 1 online resource (110 p.)
Language: english
Creator: Lee, Jongdae
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: Electrical and Computer Engineering -- Dissertations, Academic -- UF
Genre: Electrical and Computer Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: A communication protocol is a set of standard rules for data representation, signaling, authentication, and error detection required to send information over a communication channel. Historically, we have used various media as communication channels; copper wire, coaxial cable, air, power lines, and various protocols have been standardized for using those channels based on the characteristics of communication. HomePlug 1.0 protocol, the first and most commonly used technology for power line communication, is based on a Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA), where a station listens to the medium before transmission to avoid collision. Due to the contention property of CSMA/CA protocol, the performance of HomePlug 1.0 MAC becomes critical when the number of users increases, just as in 802.11 wireless MAC protocol. Another big issue in a CSMA/CA-controlled network is a hidden terminal problem which occurs because the carrier sensing range is different from the transmission range. This hidden terminal problem increases the potential collision probability which dramatically degrades the network performance. Based on the original HomePlug 1.0, we proposed an Optimal Constant Contention Window (OCCW) mechanism which dynamically selects the best fixed contention window size according to the number of contending stations in the channel. An analytical and simulation framework was used to find the optimal values of contention window size for best performance, and it is shown that using the proposed scheme results in almost constant throughput of about 80% independent of the number of nodes. Also we implemented a node estimation algorithm that each station can use independently to find the number of contending stations. The simulation results demonstrated that this estimation algorithm was well tracking the predefined actual number of nodes; moreover, it provided much better performance than the original HomePlug 1.0 protocol even under some hidden terminal conditions. Another part of research, an efficient TCP mechanism by reducing out-of-sequence packets which may happen during handover procedure in mobile IPv6 has been proposed at the end of this dissertation. A packet reordering algorithm based on the snoop protocol has been proposed to arrange the out-of-sequence packets and simulation results demonstrated that managing the packet sequence increases the overall TCP performance by reducing the packet re-transmission in Mobile IPv6 network.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Jongdae Lee.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.
Local: Adviser: Latchman, Haniph A.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022857:00001


This item has the following downloads:


Full Text

PAGE 1

PERFORMANCE ENHANCEMENT OF CSMA/CA IN POWERLINE COMMUNICATIONS UNDER HEAVY TRAFFIC AND HIDDEN NODE CONDITIONS By JONGDAE LEE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008 1

PAGE 2

2008 Jongdae Lee 2

PAGE 3

To my Parents and Family 3

PAGE 4

ACKNOWLEDGMENTS I thank my academic advisor, Dr. Haniph A. Latchman for his patient guidance, encouragement and plentiful advice while I finish my doctoral resear ch. I am also grateful to Dr. Janise McNair, Dr. Richard Newman, and Dr. Norman Fitz-Coy for serving on my advisory committee and for their helpful advice. I express my deepest gratitude to my family members who have encouraged me and been a source of power over the years. Without their en dless love, faith, and ceaseless support, I could not stand at this point. I am indebted to Dr. Chan-Do Lee, a professor in the Department of Information and Communications Engineering at Daejeon University, Korea, w ho gave me a first chance in studying abroad. Finally, I thank all LIST Lab members for th eir support and feedback on my research. It was a great pleasure to work with them. I especially thank my coworker, Dr. Kartikeya Tripathy, for his help. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........8 LIST OF FIGURES.........................................................................................................................9 ABSTRACT...................................................................................................................................11 CHAPTER 1 INTRODUCTION................................................................................................................. .13 Historical Background.......................................................................................................... ..13 Medium Access Control Protocols.........................................................................................15 Medium Sharing Techniques...........................................................................................15 Random access.........................................................................................................16 Scheduling approaches to medium...........................................................................19 Contribution and Organization...............................................................................................21 2 OVERVIEW OF WIRELESS AND POWERLINE COMMUNICATION ENVIRONMENT...................................................................................................................24 Power Line Communication Environment.............................................................................24 HomePlug 1.0 PHY.........................................................................................................25 HomePlug 1.0 MAC........................................................................................................29 Priority resolution.....................................................................................................29 Random backoff procedures.....................................................................................30 Experimental Performan ce of HomePlug 1.0 MAC........................................................31 Wireless Communication Environment..................................................................................33 MAC Performance of original 802.11 Wireless LAN.....................................................34 3 PERFORMANCE ANALYSIS OF CS MA/CA BASED MAC PROTOCOLS....................38 Binary Exponential Backoff...................................................................................................39 Dynamic Tuning Backoff.......................................................................................................41 Fast Collision Resolution...................................................................................................... ..42 Exponentially Increase Exponentially Decrease....................................................................43 5

PAGE 6

4 CONTENTION WINDOW BASED PARA METER SELECTION TO IMPROVE POWER LINE MAC EFFICIENCY FO R LARGE NUMBER OF USERS.........................46 Introduction................................................................................................................... ..........46 Constant Contention Window MAC......................................................................................49 Mathematical Analysis....................................................................................................49 Contention Window Parameterization............................................................................51 Numerical Results...........................................................................................................52 Conclusions.............................................................................................................................56 5 PERFORMANCE ENHANCEMENT OF CONSTANT CW BASED MAC PROTOCOL WITH A NODE ESTIMATION SCHEME.....................................................57 Introduction................................................................................................................... ..........57 Estimation of Number of Nodes.............................................................................................57 Estimation Strategy.........................................................................................................58 Filter Implem entation......................................................................................................59 Numerical Results for Estimation...................................................................................60 Conclusions.............................................................................................................................63 6 PERFORMANCE EVALUATION OF MODIFIED HOMPLUG 1.0 MAC PROTOCOL UNDER HIDDEN NODE CONDITIONS.......................................................64 Introduction................................................................................................................... ..........64 Hidden Node Conditions........................................................................................................66 Hidden Node Scenarios..........................................................................................................68 Simulation Results and Conclusions......................................................................................71 7 AN EFFICIENT TCP MECHANISM TO REDUCE OUT-OF-SEQUENCE PACKETS IN MOBILE IPV6................................................................................................................. .73 Introduction................................................................................................................... ..........73 Related works and problems...................................................................................................74 Standard mobile IPv6......................................................................................................74 Fast Mobile IPv6.............................................................................................................76 Improving TCP Mechanism to Reduce Disordered Packet (I-TCP)......................................79 I-TCP for movement detection........................................................................................79 I-TCP for data................................................................................................................. .79 Simulation Results............................................................................................................. .....81 Conclusions.............................................................................................................................84 8 CONCLUSIONS AND FUTURE WORK.............................................................................85 6

PAGE 7

APPENDIX A PROGRAMMING SOURCE CODES...................................................................................87 Original Homeplug 1.0 MA C Backoff Algorithm.................................................................87 The IEEE 802.11 Backoff Algorithm.....................................................................................89 Optimal Constant Contention Window based MAC..............................................................90 Estimation Algorithm for Tracking th e Actual Number of Nodes.........................................92 Proposed Optimal CW based Backoff Algorith m with Estimation Scheme under Hidden Node Conditions.................................................................................................................97 LIST OF REFERENCES.............................................................................................................106 BIOGRAPHICAL SKETCH.......................................................................................................110 7

PAGE 8

LIST OF TABLES Table page 2-1 Different priority le vels of user traffic...............................................................................29 2-2 Relationship between CW and DC as a function of BPC and priorities............................31 2-3 Pseudo code for the MAC efficiency simulation of HomePlug 1.0 protocol....................32 2-4 The IEEE 802.11 channel parameters................................................................................35 2-5 Pseudo code for the MAC efficien cy simulation of IEEE 802.11 protocol.......................36 3-1 Typical backoff algorithms summarization.......................................................................44 4-1 Pseudo code for the MAC efficiency simulation of HomePlug1.0 and proposed scheme................................................................................................................................55 5-1 Pseudo code for the estimation algorithm for figure 5-2...................................................62 8

PAGE 9

LIST OF FIGURES Figure page 1-1 Multiple access communications.......................................................................................15 1-2 Approaches to shari ng a transmission medium.................................................................16 1-3 Basic reservation system: each statio n has own minislot for reservations........................19 1-4 Examples of polling systems.............................................................................................20 2-1 Orthogonal frequency division multiplexing transceiver..................................................27 2-2 Timing sequences in HomePlug 1.0 MAC........................................................................30 2-3 The MAC efficiency of HomePlug 1.0..............................................................................32 2-4 The IEEE 802.11 DCF basic access mechanism...............................................................33 2-5 The MAC efficiency of IEEE 802.11 MAC protocol........................................................36 3-1 State diagram of backoff mechan ism of BEB: CWmin=16, CWmax = 1024...................40 3-2 Backoff mechanism of EI ED: CWmin = 16, CWmax = 1024..........................................45 4-1 State space representation of Markov ch ain model for constant contention window based HomePlug 1.0..........................................................................................................48 4-2 Comparison of analytical and simulation based performance of constant contention window based HomePlug 1.0............................................................................................51 4-3 Optimal p0 value as a function of number of stations; the different curves correspond to varying TI/ TC ratios........................................................................................................53 4-4 The p0 as a function of W and for 5 stations...................................................................53 4-5 Variation of W that gives optimal p0 as a function of number of stations, for two values of ..........................................................................................................................54 4-6 Comparison of performance of standard HomePlug 1.0 MAC and constant CW based MAC with optimal window sizes............................................................................55 5-1 Observation window for updati ng the estimate of population...........................................59 5-2 Simulation results for es timating number of nodes...........................................................60 5-3 Solution of Eq. 5-1 for pi against number of nodes...........................................................61 9

PAGE 10

5-4 Simulation results of MAC effi ciency with estimation scheme........................................61 6-1 Hidden node environment in wi reless local area network.................................................66 6-2 Hidden node environment in power line network..............................................................67 6-3 The MAC efficiency under various hidden terminal conditions.......................................69 7-1 Standard mobile IPv6 handover procedure........................................................................74 7-2 Out-of-sequence packet problem in a mobile IPv6 network.............................................76 7-3 Fast handover for mobile IPv6...........................................................................................77 7-4 Out-of-sequence packet problem in FMIPv6 network.......................................................78 7-5 Proposed EAP-MIPv6 signaling flow................................................................................79 7-6 Proposed I-TCP scheme for data.......................................................................................80 7-7 Simulation network scenario..............................................................................................81 7-8 Packet sequence in stan dard MIPv6 during handover.......................................................82 7-9 Packet sequence in FMIPv6 during handover...................................................................83 7-10 Packet sequence in proposed I-TCP during handover.......................................................83 10

PAGE 11

Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy PERFORMANCE ENHANCEMENT OF CSMA/CA IN POWERLINE COMMUNICATIONS UNDER HEAVY TRAFFIC AND HIDDEN NODE CONDITIONS By Jongdae Lee December 2008 Chair: Haniph A. Latchman Major: Electrical and Computer Engineering A communication protocol is a set of standard rules for data representation, signaling, authentication, and error detec tion required to send informati on over a communication channel. Historically, we have used various media as communication channels; c opper wire, coaxial cable, air, power lines, and various protocols have been standardized for using those channels based on the characteristics of communication. HomePlug 1.0 protocol, the first and most commonly used technology for power line communication, is based on a Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA), where a station listens to the medi um before transmission to avoid collision. Due to the contention property of CSMA/CA protocol, the performan ce of HomePlug 1.0 MAC becomes critical when the numbe r of users increases, just as in 802.11 wireless MAC protocol. Another big issue in a CSMA/CA-controlled network is a hidden terminal problem which occurs because the carrier sensing range is diffe rent from the transmission range. This hidden terminal problem increases the potential collis ion probability which dramatically degrades the network performance. Based on the original HomePlug 1.0, we proposed an Optimal Constant Contention Window (OCCW) mechanism which dynamically se lects the best fixed contention window size 11

PAGE 12

according to the number of contending stations in the channel. An analytical and simulation framework was used to find the optimal values of contention window size for best performance, and it is shown that using the proposed scheme results in almost cons tant throughput of about 80% independent of th e number of nodes. Also we implemented a node estimation algorith m that each station can use independently to find the number of contending stations. Th e simulation results demonstrated that this estimation algorithm was well tracking the predef ined actual number of nodes; moreover, it provided much better performance than the original HomePlug 1.0 protocol even under some hidden terminal conditions. Another part of research, an efficient TC P mechanism by reducing out-of-sequence packets which may happen during handover procedure in m obile IPv6 has been proposed at the end of this dissertation. A packet reordering algorithm based on the s noop protocol has been proposed to arrange the out-of-sequence packets and simu lation results demonstrated that managing the packet sequence increases the overall TCP perf ormance by reducing the packet re-transmission in Mobile IPv6 network. 12

PAGE 13

CHAPTER 1 INTRODUCTION Historical Background Power Line Communication (PLC) systems, which have been around since the 1950s but were never seriously thought of as a communica tion method due to the low speed and high cost for development, are now gradually spreading to local area networks and broadband over power lines as a realistic and practic al means of communication. Phrase s such as "Inv isible Wiring Communication" and "No New Wi re Communication" were attr active enough to spread PLC systems. While many have attempted to use the power line as a communication medium in the past, it has not lived up to expectations, earning a reput ation for questionable reliability. The fact is that the power line is an extremely difficult and noisy communication medium, characterized by several unpredictable and strong form s of interference. Every applia nce that is connected to an outlet contributes line interfer ence, which can be approximated as Additive White Gaussian Noise (AWGN). Dimmer switches, motorized electrical appliances, and computers also introduce pulse interference into the communications environment, making it a perilous channel for non-robust communications. Fast fading also corrupts power line comm unication channels, which have non-flat frequency responses and suffer from unpredictabl e jamming and the presence of carrier wave signal interference. This makes it exceedingl y difficult to ensure reliable power line communication and help explain the difficulties technology has had in delivering reliable power line communication. Because of these harsh cond itions, it was natural that people tried to find another way to get affordable and portable devi ces for home, entertainment and personal use to 13

PAGE 14

communicate with each other on a multimedia and digital platform in the local area network environment. With the advantage of simplicity to implement cable modem and various forms of digital subscriber lines (DSL) were developed first, but their low download and upload speed was insufficient for high data traffic communicati on. Another approach is to deploy an IEEE 802.11 wireless local area network (WLAN) with wirele ss modems in each device connecting to one or more wireless hubs (infrastructur e-based) or to each other (a d-hoc based). The wireless option, with data rates now up to 54Mbps and the big advant age of flexibility, is certainly viable except that a dedicated wired infrastructure connecti ng multiple access points is required to cover the entire home. Power lines, being ubiquitously deployed as a wire-line network for carrying electrical power, are then the obvious choice as the medi um for communication am ongst the plethora of home-based and personal devices. They offer the convenience of already being there and having outlets in almost all locations in a household for easy access. Fu rther, devices can easily obtain electric power if they are deployed on PLC sy stems, while wireless mobile devices rely on batteries and have difficulty maintaining c ontinuous electric power. HomePlug 1.0 [1], standardized in 2001, is one of the most popular power line communicati on technologies, and it supports up to 14Mbps transmission rate using power lines. However, the PLC systems are not free of problems. The PLC channel is notorious for electric noise and interference, as well as channe l variability depending on the appliances that are in use from time to time. Even then, with th e support of advanced modulation and channel coding techniques, e.g., Orthogonal Frequency Division Multiplexing (OFDM) and Forward Error Correction (FEC) [2], the present versi on of the HomePlug 1.0 has been shown to out14

PAGE 15

perform the traditional IEEE 802.11 a/b in many field tests of connectivity, throughput and link stability [3]. Medium Access Control Protocols The Media Access Control (MAC) data communi cation protocol sub-layer, also known as the Medium Access Control, is a sub-layer of the Data Link Laye r specified in the seven-layer OSI model (layer 2). It provide s addressing and channel access control mechanisms that make it possible for several terminals or network nodes to communicate within a multipoint network, typically a local area network (LAN) or metropolitan area network (MAN). The MAC sub-layer acts as an interface betw een the Logical Link Control (LLC) sub-layer and the network's physical layer. The MAC laye r emulates a full-duplex logical communication channel in a multipoint network. This channel may provide unicast, multicast or broadcast communication service. Figure 1-1 Multiple access communications Medium Sharing Techniques Figure 1-1 shows a generic multiple access comm unications situation in which a number of user stations share a transmission medium. The transmission medium is broadcast in nature, and so all the other stations that are attached to the medium can hear that transmission from any 15

PAGE 16

given station. When two or more stations transmit simultaneously, their signals will collide and interfere with each other. There are two broad categories of schemes for sharing a transmission medium. The one category involves a static and collision-free sharing of the medium, channelization schemes because they involve the partitioning of the me dium into separate channels that are then dedicated to particular users. Channelization techniques are suita ble when stations generate a steady stream of information th at makes efficient use of the dedicated channel. The other category involves sharing of the medium dynamically on a per frame basis that is better matched to situations where bursty traffic condition. We refer to this category as MAC schemes. The primary function of medium access control is to mi nimize or eliminate the incidence of collisions to achieve a reasonable utilizati on of the medium. Figure 1-2 summ arizes the various approaches to sharing a transmi ssion medium [4]. Medium sharing techniques Static channelization Dynamic medium access control Scheduling Random access Figure 1-2 Approaches to shar ing a transmission medium Random access ALOHA is the first radio network used this random access scheme. The University of Hawaii needed a means to interc onnect terminals at campuses located on different islands to the host computer on the main campus. A radio transmitter is attached to the terminals, and 16

PAGE 17

messages are transmitted as soon as they become available, thus producing the smallest possible delay. From time to time frame transmissions will collide, but these can be treated as transmission errors, and recovery can take place by retransmission. When traffic is very light, the probability of collision is very small, and so retr ansmissions need to be carried out infrequently. Consequently, under light traffic the frame transfer delay will be low. There is a significant difference between normal transmission errors and those that are due to frame collisions. Transmission errors that are due to noise affect only a single station. On the other hand, in the frame collisions more than on e station is involved, and hence more than one retransmission is necessary. This interaction by se veral stations produces a positive feedback that can trigger additional collisions. For example, if the stations use the same time-out values and schedule their retransmissions in th e same way, then their future retransmissions will also collide. For this reason, the ALOHA scheme requires stations to use a backoff algorithm, which typically chooses a random number in a certain retransm issions time interval. This randomization is intended to spread out the retransmissions a nd reduce the likelihood of additional collisions between the stations. More detail s about the backoff algorithm have been described in Chapter 2. The low maximum throughput of the ALOHA schemes is due to the wastage of transmission bandwidth because of frame collis ions. This wastage can be reduced by avoiding transmissions that are certain to cause collisi ons. By sensing the medium for the presence of a carrier signal from other stations, a stati on can determine whether there is an ongoing transmission. The class of carrier sensing multiple access MAC schemes uses this strategy. CSMA schemes differ according to the behavior of stations that have a frame to transmit when the channel is busy. In 1-Persistent CSMA stations with a frame to transmit sense the channel. If the channel is busy, they sense th e channel continuously, wa iting until the channel 17

PAGE 18

becomes idle. As soon as the channel is sensed id le, they transmit their frames. Consequently if more than one station is waiting, a collision will occur. In a sense, in 1-Persistent CSMA stations act in a greedy fashion, attempting to access the medium as soon as possible. As a result, it has a relatively high collision rate. Non-Persistent CSMA attempts to reduce the in cidence of collisions. Stations with a frame to transmit sense the channel. If the channel is busy, the stations imme diately run the backoff algorithm and reschedule a future resensing time. If the channel is idle, the stations transmit. By immediately rescheduling a resens ing time and not persisting, th e incidence of collisions is reduced relative to 1-Persistent CSMA. This imme diate rescheduling also results in longer delays than are found in 1-Persistent CSMA. The class of p-Persistent CS MA schemes combines elements of the above two schemes. Stations with a frame to transmit sense the channel, and if the channel is busy, they persist with sensing until the channel become s idle. If the channel is idle, the following occurs: with probability p, the station transmits its frame; with probability 1-p the station decides to wait an additional propagation delay before gain sensing the channel. This behavior is intended to spread out the transmission attempts by th e stations that have been wa iting for a transmission to be completed and hence to increase the likelihood th at a waiting station su ccessfully seizes the medium [4]. Carrier sense with multiple access and collis ion avoidance (CSMA/CA) is the MAC layer mechanism used by IEEE 802.11 WLANs and HomePl ug 1.0 PLC networks. Carrier sense with multiple access and collision detection (CSMA/ CD) is a well-studied technique in IEEE 802.x wired LANs. This technique cannot be used in the context of WLANs effectively because the error rate in WLANs is much higher and allowing collisions will lead to a drastic reduction in 18

PAGE 19

throughput. Moreover, detecting collisions in the wireless medium is not always possible. The PLC network environment is very similar to thes e WLANs, especially ad hoc mode of WLANs, in which communications are not controlled by Access Point. Therefore, one of collision avoidance techniques is adapted to both WLANs and PLC networks. Figure 1-3 Basic reservation system: each st ation has own minislot for reservations Scheduling approaches to medium The scheduling approaches attempt to produ ce an orderly access to the transmission medium. Polling systems as a special form of reservation systems are good example of this approach. Figure 1-3 shows a basic reservation system. The stations take turns transmitting a single frame at the full rate R bps, and the tran smissions from the stations are organized into cycles that can be variable in length. Each cycle begins with a reservation interval. In the simplest case the reservation interval consists of M minislots, one minislot per station. Stations use their corresponding minislot to indicate that they have a fram e to transmit in a corresponding cycle. The stations announce thei r intention to transmit a frame by broadcasting their reservation bit during the appropriate minislot. By listening to the reservation interval, the stations can determine the order of frame transmissions in the corresponding cycle. The length of the cycle 19

PAGE 20

will then correspond to the number of stations that have a frame to transmit. Note that variablelength frames can be handled if the reservati on message includes frame-length information. In polling system, stations take turns accessing the medium. At any given time only one of the stations has the right to transmit into the me dium. When a station is done transmitting, some mechanism is used to pass the right to transmit to another station [4]. Figure 1-4 Examples of polling systems There are different ways for passing the right to transmit from station to station. Figure 14A shows the situation in which M stations communicate with a host computer. The system consists of an outbound line in which information is transmitted from the host computer to the stations and an inbound line that must be shared with the M stations. The inbound line is a shared medium that requires a medium access control to coordinate the transmissions from the stations to the host computer. The techni que developed for this system involves the host computer acting 20

PAGE 21

as a central controller that issues control messages to coordinate the transmissions from the stations. The central controller sends a polling me ssage to a particular station. When polled, the station sends its inbound frames and indicates the completion of its transmission through a goahead message. The central controller might po ll the stations in round-robin fashion, or according to some other pre-determined order. Figure 1-4B shows another situat ion where polling can be use d. Here the central controller may use radio transmissions in a certain frequency band to transmit outbound frames, and stations may share a different frequency band to transmit inbound frames. This technique is called the frequency-division duplex (FDD) approach. Again the centr al controller can coordinate transmissions on the inbound channel by issuing polling message s. Another variation of Figure 1-4B involves having inbound and out bound transmissions share one frequency band. This is time-division duplex (TDD) approach. In this case we would have an alternation between transmissions from the central controller and transmissions from polled stations. Figure 1-4C shows a situation wh ere polling is used without a central controller. In this particular example we assume that the stations have developed a polling order list, using some protocol. We also assume that a ll stations can receive the transm issions from all other stations. After a station is done transmitting, it is respons ible for sending a polling message to the next station in the polling list. Contribution and Organization Thanks to the development of advanced tec hnology result from many researchers efforts to improve the overall performance of HomePl ug 1.0 standard, the HomePlug 1.0 protocol could overcome lots of restrictions, and even an advanced version has been developed by some commercial companies. However, just as in IEEE 802.11, HomePlug 1.0 MAC protocol still has a fatal flaw that the throughput performance beco mes critical when the number of users increases 21

PAGE 22

because it is CSMA/CA based and extends th e random backoff algorith m of IEEE 802.11. This results from Binary Exponential Rand om Backoff property in CSMA/CA. As the contention nature of this protocol, one of the main issues for the research is to improve network performance that comes from the collision resolution mechanism under various traffic scenarios. For example, when there are few active nodes in the system, the transmission probability of each node should be increased to redu ce the average number of idle time slots. But if the traffic load is heavy, th e rate should be decreased in or der to avoid more collisions. Another kind of algorithms to increase the system performance is the improvement of transmission efficiency. One direct scheme is to increase the physical transmission rate of current protocol. This research work has been car ried out by IEEE 802.11 WG. But Xiao [5] shows a theoretical throughput limit exists due to the ov erhead of MAC and PHY layer in current 802.11 DCF. Therefore solely increasing the transmi ssion rate would not be very helpful. For this reason, our research has been focused on the former approaches and developed optimal constant contention window based MAC protocol, a modificatio n of existing HomePlug 1.0 MAC protocol, and applied this modified one to power line communication network environment. This protocol is shown to significantly enhance the MAC performance under saturation conditions. An analytical and simulation framework has been used to tune the modified protocol for best performance. This requires the knowledge of the active number of users on the network. Chapter 2 presents an access scheme of CS MA/CA based MAC protocol both in wireless and power line communication networks and the MAC efficiencies are shown by simulations. In Chapter 3, we gives a related work for the perf ormance enhancement of carrier sensing multiple access MAC protocols and discuss on the problems in those protocols. Chapter 4 presents a 22

PAGE 23

modified MAC access scheme in power line co mmunications as a solution to keep high performance in a massive network circumstances (a large number of users in the network). In Chapter 5, independent node estimation scheme wh ich should be premised to use the optimal contention window based MAC access scheme is pr oposed. In Chapter 6, a computer simulation of this modified HomePlug 1.0 MAC protocol under the circumstances of hidden terminal problems in the network is proposed. In Chapter 7, another part of research in this dissertation, an efficient TCP mechanism by reducing out-ofsequence packets in mobile IPv6 is proposed. Chapter 8 concludes the dissertation al ong with the plan of future work. 23

PAGE 24

CHAPTER 2 OVERVIEW OF WIRELESS AND POWERLINE COMMUNICATION ENVIRONMENT CSMA/CA is a modification of pure Carrier Sense Multiple Access (CSMA). Collision avoidance is used to improve the performance of CSMA by attempting to be less "greedy" on the channel. If the channel is sensed busy before tr ansmission then the transmission is deferred for a "random" interval. This reduces the probability of collisions on the channel. CSMA/CA is used where CSMA/CD cannot be implemented due to the nature of the channel. CSMA/CA is used in 802.11 based wireless LANs. One of the problems of wireless LANs is that it is not possible to listen while sending; therefore collis ion detection is not possible. In case of power line communi cation networks, the most popular wired MAC protocol, CSMA/CD, could be applied, but the large variation in noise on the power line makes collision detection very hard. Because of this harsh ch annel conditions, some have applied the CSMA/CA protocol as suggested in IEEE 802.11 to the power line network. In this chapter, we will analyze the perfor mance of CSMA/CA based MAC protocol in two different kinds of network environment, wire less and power line communication networks and find the problems of this protocol. Power Line Communication Environment PLC transmits data over an existing high-volta ge power line instead of requiring dedicated cabling. PLC can thus provide an inexpensive solution for transmitting data in a pre-wired location. Power lines were orig inally devised for distributi ng electrical po wer using the frequency range of about 50-60 Hz. The use of this medium for high speed communications presents some technical challenging problems. Electrical noise from appliances and the uncontrolled nature of the wiring resu lt in severe signal distortions. 24

PAGE 25

The PLC channel is made up of different conductor types; ther efore a variety of characteristic impedances will be encountered. Further, the network terminal impedance will tend to vary with frequency and time as the c onsumers load pattern and load types vary. Impedance mismatch causes multi-path effects resulting in deep notches at certain configuration dependent frequencies. These channel imperfecti ons make signal transmission over a power line very difficult [6]. Reliable data communication over this hostile medium requires powerful Forward Error Correction (FEC) coding, interleaving, error det ection and Automatic Repeat Request (ARQ) techniques, along with appropria te modulation schemes as well as a robust Medium Access Control (MAC) protocol. These obstacles were a sever factor that makes transmitting data over electrical power lines not big issued at the ear ly development. However, recent advances in communication and modulation methodologies such as adaptive digital signal processing and variable coding schemes for error controls have spawned novel media access control (MAC) and physical layer (PHY) protocols capable of s upporting power line communication networks at speeds comparable to wired local area networks (LANs). HomePlug 1.0 PHY To overcome the hostile PLC environment, Orthogonal Frequency Division Multiplexing (OFDM) with a Cyclic Prefix (CP) was adopted by the HomePlug 1.0 PLC standard. Using OFDM has many benefits as follows; it exhibits excellent mitigation of the effects of timedispersion, provides excellent Inter-Channel Inte rference (ICI) performance, and is good at minimizing the effect of in-b and narrowband interference. OFDM is a discrete multi-tone technology in which numerous signals of different frequencies, called carriers, are combined to form a single signal for transmission. Prior to combining, each carrier is first "phase shifted," or modulated, for the purpose of representing 25

PAGE 26

data bits. By modulating data bits on individual signals prior to combining them, many data bits can be transmitted over a small amount of time. This approach is used by HomePlug 1.0. OFDM is well know in the literature a nd in industry [1, 7]. To obtain high spectral efficiency the frequency response of the subcarriers are overlapping and orthogona l. Each narrowband subcarrier can be modulated using various modulation formats. By choosing the subcarrier spacing to be small the channel transfer functi on reduces to a simple constant within the bandwidth of each subcarrier. In this way, a freque ncy selective channel is divided into many flat fading subchannels, which eliminates the need for sophisticated equalizer s. The followings are the advantages and disadvantages of OFDM: Advantages: Robust to narrow-band co-channel interference High spectral efficiency Robust to inter-symbol interference and fading caused by multipath propagation Flexible and can be made adaptive; differe nt modulation schemes for subcarriers, bit loading, adaptable bandwidt h/data rates are possible Disadvantages: Sensitive to Doppler shift Sensitive to frequency synchronization problems Inefficient transmitter power consumption A more advanced technique cal led bit-loading allows use of different modulation and coding schemes for each sub-carri er. In either case, OFDM can adapt bandwidth/data rates according to channel conditions. 26

PAGE 27

Figure 2-1 Orthogonal frequency di vision multiplexing transceiver 27

PAGE 28

Figure 2-1 shows an overall block diagram of the OFDM transceiver. Before forming the OFDM symbol in the Analog Front End (AFE), data are scrambled, RS-encoded, convolution encoded, punctured, then interleaved on the transmitter. These processes will be discussed in more detail below. The AFE consists of a constellation mapping block, an IFFT block, a preamble block, a CP block, and a raised cosine (RC) block. The mapping block groups data bits and maps them onto a constellation point of the modulation method; it selects the type of modulation and the carriers to be used in the I FFT block, as specified by the tone map and tone mask. The IFFT block modulates the constellation points onto the carrier waveforms (in discrete time), while the preamble block inserts the pream ble. The CP is added by the CP block, and RC shaping is employed to reduce out-of-channel energy. The bandwidth in HomePlug 1.0 can vary from 1 Mbps to 14 Mbps practically continuously according to the cannel conditi ons. Active HomePlug 1.0 nodes perform channel estimation at least once every 5 seconds. This f eature allows the PLC network to maximize its data rate adaptively. A preamble and frame control form delimiters used for synchronization and for control. The frame control of start of frame, end of fram e, and response delimiters all include delimiter type, and contention control information. In the st art of frame delimiter, the frame control field includes the tone map information needed by the rece iver to decode the rest of the frame, and a length field. The end of frame delimiter contains pr iority information used for contention control. Response delimiters contain information that allo ws a sender to verify that the response was indeed sent in response to the frame it just transmitted. An end of frame gap (EFG) of 1.5 s is inserted between the frames frame check seque nce (FCS) and the end delimiter to allow for processing. 28

PAGE 29

HomePlug 1.0 MAC The HomePlug 1.0 Medium Access Control (M AC) protocol is a modified CSMA/CA (Carrier Sense Multiple Access / Collision Avoidance) protocol with priority signaling. HomePlug 1.0 devices operate in an ad hoc mode in the sense that devices communicate with each other freely, without any centralized coordination. The HomePlug 1.0 standard uses different terms and stages for inter-frame spacing and for the contention windows comparing to 802.11. The RIFS shown in the Figure 2-2 is Response Inter-Frame Spacing. Unlike 802.11, there is no SIFS (Short Inter-Frame Spacing) between continued frames. Rather, a frame control bit is used to indicate the desire of a station to continue to send data, allowing preemption only by higher prio rity traffic. The spacing between the last frame and the incoming frame is CIFS (Contention Window Inter-Frame Spacing). Priority resolution In HomePlug 1.0, quality of service (QoS) is provided by differentiating user traffic into 4 priority levels: CA0, CA1, CA2 and CA3. The first two are lower priority traffic, and CA2 and CA3 have higher priorities. The higher the prior ity of the data packet, the earlier it gets to contend for the channel. Table 2-1. Different priority levels of user traffic Traffic Classes PRS0 PRS1 CA3 (Voice) CA2 (Audio/Video) CA1 (Bulk transfers/Background) CA0 (Best effort) Table 2-1 shows the different priority levels of user traffic used in current HomePlug 1.0 MAC protocol. Before channel contention begins, HomePlug 1.0 arranges for the resolution of the priorities of contending traffi c by introducing two priority resolution slots (PRSs), PRS0 and 29

PAGE 30

PRS1. All stations having frames to transmit send signals in these slots, in accordance with the priorities of their data. Stations with CA3 data can send signals in PRS0 and PRS1. Stations with CA2 data can send signals in PRS0. Stations with CA1 data can send signals in PRS1. Stations with CA0 data can not send signals in either of these slots. The two PRSs inform all the stations of the priorities of other stat ions hopeful of contending for th e channel, and those with lower priorities defer to those with higher priorities. Once the stations th at will be in the contention are decided, they begin their ra ndom backoff procedures. This is shown in Figure 2-2. The station that wins the contention transmits its data, and the response is received after a time equal to Response Interframe Space (RFIS) has elapsed. After the response of a successful transmission is receiv ed, the next priority resolution session begins after Contention Inte rframe Space (CIFS). Figure 2-2 Timing sequences in HomePlug 1.0 MAC Random backoff procedures The backoff algorithm of HomePlug 1.0 is diffe rent from that of IEEE 802.11 in that HomePlug 1.0 uses three counters: backoff procedure counter (BPC), deferral counter (DC) and backoff counter (BC). BPC and BC represent the number of retransmissions (backoff stage, corresponding to the s value in 802.11 [8]) and the random backoff time (corresponding to the b value in 802.11 [8]), respectively. In HomePlug 1.0, DC is newly in troduced to roughly estimate the number of contending stations. Every sta tion starts its contention for the channel by initializing the value of BPC to 0, and choos ing BC randomly between 0 and CW0-1, where 30

PAGE 31

CW0 denotes the initial contenti on window (CW) size. The value of DC is set depending on the value of BPC: every time BPC changes, DC (a nd BC) are updated with respect to Table1. If a slot is sensed idle and BC is not zero yet, BC is decreased by one while DC and BPC are fixed (if BC becomes zero, the station transmits). If a slot is sensed busy, both BC and DC are decreased by one at the end of that busy sl ot. If DC then becomes less than zero, the BPC is updated to the next higher value, and both DC and BC are rein itialized according to th e new BPC value (Table 2-1). In that case, BC is chosen randomly from 0 to CWi-1, wher e CWi is the new CW for BPC i. The same happens in the case of a collision also. In case of a success, the BPC is set to its minimum value, and DC and BPC are set accordingly. Table 2-2 Relationship between CW and DC as a function of BPC and priorities Priorities CA3, CA2 Priorities CA1, CA0 BPC = 0 DC = 0 CW( W0)= 8 DC = 0 CW( W0)= 8 BPC = 1 DC = 1 CW( W1)= 16 DC = 1 CW( W1)= 16 BPC = 2 DC = 3 CW( W2)= 16 DC = 3 CW( W2)= 32 BPC > 2 DC = 15 CW( W3, ..)= 32 DC = 15 CW(W3, )= 64 Experimental Performance of HomePlug 1.0 MAC Based on the concept of the above back off algorithm, some have already showed the MAC performance under slightly different kinds of assumptions. We also worked out a detailed analysis for the MAC performance of HomePl ug 1.0 under the following assumptions: there are n stations having frames for transmission; all sta tions have frames with the same priority; each station has a frame immediately after the successful completion of a frame transmission; the transmission channel is ideal collisions are de tected only when there ar e two or more stations whose back off counter reach to zero at some point so they send their packets. Under these assumptions, we construct a system model and got the simulation result as shown in the Figure 2-3. 31

PAGE 32

Table 2-3 Pseudo code for the MAC efficiency simulation of HomePlug 1.0 protocol At successful transmission At collision At idle if (success) node's BPC = 1st stage of BPC; node's DC = 1st stage of DC; node's BC = randomly select (0, CW0); others' DC = DC 1; others' BC = BC 1; if (others' DC < 0) their BPC = next stage of BPC; their DC = new DC; their BC = randomly select (0, newCW-1); end end if (collide) node's BPC = next stage of BPC; node's DC = new DC; node's BC = randomly select (0, newCW-1); others' DC = DC 1; others' BC = BC 1; if (others' DC < 0) their BPC = next stage of BPC; their DC = new DC; their BC = randomly select (0, newCW-1); end end if (idle) all node's BC = BC 1; end 10 20 30 40 50 60 70 80 90 100 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Number of NodesMAC Efficiency CA1, CA0 CA3, CA2 Figure 2-3 The MAC efficiency of HomePlug 1.0 32

PAGE 33

Table 2-3 is the pseudo code for the MAC ef ficiency simulation of HomePlug 1.0 protocol for PLC network. We ran the simulation with 2 di fferent priorities leve ls. Based on the results shown in the Figure 2-3, we can see that the MAC throughput decreases as number of stations, n, increases because the probability of collision increases. For a fixed value of n, MAC throughput of the low priority group is larger than that of the high priority gr oup, since CW of the low priority group is larger than that of the high priority group. However, the most important issue in this result is that the MAC throughput remarkab ly decreases as the number of node increase. Wireless Communication Environment Figure 2-4 The IEEE 802.11 DCF basic access mechanism Wireless LAN also has some restrictions to use CSMA/CD as a MAC layer protocol. The first reason is that it is diffi cult to detect collisions in a radio environment. Because the transmitted power would overwhelm the received power at the same station, it is not possible to abort transmissions that collide. A second reason is that the ra dio environment is not as well controlled as a wired broadcast medium, and transmissions from users in other LANs can interfere with the operation of CSMA-CD. A third reason is that radio LANs are subject to the hidden-station problem that occurs when two stat ions attempt to transmit to a station that is located between them. This condition will result in the transmissions from the two stations proceeding and colliding at the intermediate station. The CSMA/CA medium access control was 33

PAGE 34

developed to prevent this type of collision. In this chapter, we will briefly talk about the original 802.11 MAC backoff algorithm and some modifi ed schemes for better performance. MAC Performance of original 802.11 Wireless LAN To access the medium, IEEE 802.11 employs a CSMA/CA MAC protocol with binary exponential backoff, called distri buted coordination func tion (DCF) [9]. As we described at the beginning of chapter 2, this DCF basic access is very similar to that of HomePlug 1.0 MAC protocol because HomePlug 1.0 MAC protocol is a modified CSMA /CA protocol with priority signaling. The DCF basic access method is briefly summarized as follows. A station monitors the channel until an idle period equal to a distribu ted inter-frame space (DIFS) is detected before initiating transmission of a data frame. After sensing an idle DIFS, the station generates a random backoff interval for an additional deferral time before transmitting. The backoff time counter is decreased by one slot tim e as long as the channel is sensed idle, frozen when there is a busy state on a channel (that is, another station is transmitting its packet), and resumed when the medium is sensed idle again for more than a DIFS. A station transmits its frame when the backoff counter reaches zero. At each frame tr ansmission, the backoff time is randomly chosen in the range (0, CW 1). The value CW is called the contention window. For each frame transmission, the CW takes an initial value CW0 and it doubles after each unsuccessful transmission, up to a predefined maximum of CWmax. The CW remains at CWmax for the remaining attempts. During the backoff procedure, each attempt to transmit a frame is called a backoff stage. 34

PAGE 35

Table 2-4. The IEEE 802.11 channel parameters Parameters Size Packet Size 3274 bits Slot time 20 s SIFS 10 s DIFS 50 s ACT 6 s Channel Bit Rate 1 Mbps Upon successful reception of a data frame, the destination station transmits an ACK immediately following a short inter-frame space (SI FS) time. If not receiving the ACK within a predetermined ACK_timeout or detecting the tran smission of a different frame on the channel, the transmitter will reschedule the frame transmission according to the backoff rules. Figure 2-4 is a simple diagram showing this 802.11 DCF basi c access mechanism. The standard defines an additional mechanism of RTS/CTS control handshake to be optionally used in the case a packet exceeds a specified length. Specifying th at special short request to send (RTS) and clear to send (CTS) frames should be sent prior to the transmi ssion of the actual data frame, it is incorporated to improve the system throughput by shortening the duration of the collisions such as taking place in hidden terminal problem. Based on this 802.11 DCF basic access mechanism, we implemented a simulation under the following assu mptions: the ideal channel condition and the saturation condition. Under the idea l channel condition, we did not take into consideration any loss of data packets sent by a station even c ontrol packets and hidden terminal problems. In saturation conditions, each stati on has immediately a packet available for transmission, after the completion of each successful transmission. Figur e 2-5 below is the simulation result of the IEEE 802.11 MAC protocol. We run the simulati on with two different kinds of maximum 35

PAGE 36

contention window sizes (CWmin=32, CWmax=256 and CWmin=32, CWmax=1024) and other parameters specified by the 802.11 standard [9]. Those are reported in the above table. Table 2-5 Pseudo code for the MAC efficiency simulation of IEEE 802.11 protocol At successful transmission At collision At idle MinCW=32; MaxCW=256, 1024; Backoff range = [0,CW]; if (success) node's CW = MinCW; node's BC = randomly select (0, MinCW); end MinCW=32; MaxCW=256, 1024; Backoff range = [0,CW]; if (collision), node's CW = min(2*currentCW, MaxCW); node's BC = randomly select (0, newCW); end MinCW=32; MaxCW=256, 1024; Backoff range = [0,CW]; node's BC = BC-1; 0 10 20 30 40 50 60 70 80 90 100 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Number of NodesMAC Efficiency CWmin=32, CWmax=1024 CWmin=32, CWmax=256 Figure 2-5 The MAC efficiency of IEEE 802.11 MAC protocol We used the following concept to calculate MAC efficient and detailed mathematical equations are derived in chapter 3. 36

PAGE 37

TotalSuccessfulTransmissionsFrameTime Efficiency TotalNumberofSlotsSlotTime As we can see on the Figure 2-5, the MAC Effi ciency for the basic access scheme strongly depends on the number of stations in the network. Of course we can get slightly different values of MAC Efficiency by changing some parameters such as the minimum and maximum values of contention window size, packet length or slot time. However, the main focus of the result in this simulation is that the MAC Efficiency is decreas es as the network size increase. This is very similar to that of HomePlug 1.0 MAC protoc ol in power line communication networks. 37

PAGE 38

CHAPTER 3 PERFORMANCE ANALYSIS OF CSMA/CA BASED MAC PROTOCOLS Being a simple and elegant random access protoc ol, the p-persistent CSMA/CA protocol is first analyzed by Kleinrock and Tobagi with th e S-G technique to obtain the throughput-delay characteristics for infinite stati ons in [10]. Its performance in a network with finite number of stations is studied by the same authors in [ 11], using the technique of embedded Markov chain, the computational complexity of which become s inhibitive for systems with large station population. After these two classic papers, there is a long time in which the analysis of this protocol is of very limited interest to researchers, until it is studied again much later by Cali et al. when they use p-persistent CSMA/CA as an alternative version of the IEEE 802.11 DCF MAC protocol in [12, 13]. In contrast, the wide deployment of IEEE 802.11 DCF based wireless LANs (Wi-Fi networks) has attracted great interests from researchers. There are numerous works on the performance study of DCF in terms of throughput and average frame service time, especially for the saturation case. A well-known one on this topic is first reported in [14] and further detailed in [8] by Bianchi. The BEB procedure is modeled as a discrete time two-dimensional Markov chain. The assumption of constant conditional collision probability of frames regardless of the transmission history is first adopted in these pape rs and later justified th eoretically by Sharma et al. in [15] by relating it to the mean-field models in statistical physics. A direct extension of the Markov model in [8] is proposed by Wu et al. in [16] and Chatzimisios et al. in [17] to take into consideration the retransmission limit for each frame as specified in the standard. Along this line, many papers modify the basi c Markov model proposed in [8] to consider some other details of the DCF [18-22], or adap t the Markov model to other backoff variants [2325]. Cali et al. analyze the p-persistent ve rsion of IEEE 802.11 DCF in [12, 13], using the 38

PAGE 39

concept of virtual transmission time of a frame. Based on simulation results, they conclude that the maximum throughput is achieved when the average channel idle time is equal to the average time for collisions, which is further justified an alytically in [26]. Tay and Chua analyze the protocol behavior from the long term average po int of view in [27]. By computing the average rates of transmission, collision and successful tran smission, and utilizing the condition that all these rates sum up to one, the saturation through put is obtained. Based on the analysis, some insights on how to achieve the optimal performance by adjusting the protocol parameters according to the number of stations are also give n in the paper. Following this direction, Kumar et al. represent the average transmission attempt rate as the ratio of average transmission times over the average backoff slots, and calculate it us ing the fixed point technique [28]. It is also pointed out in this paper that by using the fixed point equations, the analysis of the Markov chain in [8] is in fact not necessary. Medepalli and To bagi extend further this approach to compute the average cycle time of a frame and the saturati on throughput in [29]. Th e only paper adopts the equilibrium point analysis is [30] by Wang et al., which requires a computational effort comparable to that of the Markov chain based models. Binary Exponential Backoff In distributed multiple access, a simple yet effective random backoff algorithm is widely used to avoid collisions. In particular, the binary exponential backoff algorithm adjusts the contention window size dynamically in react to collision intensity. Such an algorithm is embedded in the IEEE 802.11 Distributed Coordination Function (DCF). Before an attempt of data transmission is made, a station senses the channel to determine whether it is idle. If the medium is sensed idle throughout a specified ti me interval, called the distributed inter-frame space (DIFS), the station is allowed to transmit. If the medium is sensed busy, the transmission is deferred until the ongoing transmission terminates. 39

PAGE 40

Figure 3-1 State diagram of backoff m echanism of BEB: CW min=16, CWmax = 1024 A slotted binary exponential backoff procedure takes place at this point: a random backoff interval value is uniformly chosen from 0 to CW -1 and used to initialize the backoff timer, where CW is the current contention window size. The backoff timer keeps running as long as the channel is sensed idle, paused when data transmi ssion (initiated by other st ations) is in progress, and resumed when the channel is sensed idle again for more than DIFS. The time immediately following an idle DIFS is slotted, with each slot e qual to the time needed for any station to detect the transmission of a frame (in the IEEE 802.11 term, MAC Service Data Unit (MSDU)) from any other station. When the backoff timer expires, the station attempts to transmit a data frame at the beginning of next slot. Finally, if the da ta frame is successfully received, the receiver transmits an acknowledgment frame after a specif ied interval, called the short inter-frame space (SIFS), which is less than DIFS. If an acknow ledgment is not received, the data frame is presumed to be lost, and a retransmission is sche duled. The value of CW is set to CWmin in the first transmission attempt, and is doubled at each retransmission up to a pre-determined value CWmax. Retransmissions for the same data fram e can be made up to a pre-determined retry limit, L, times. Beyond that, the pending frame will be dropped. However, this BEB algorithm 40

PAGE 41

suffers from fairness issues resulted from its immediate window resetting which favors the last succeeding node. Dynamic Tuning Backoff Cali et. al([12]) derive the average size of the contention window that maximizes the aggregate throughput under the assu mption that all stations have the same average contention window size of transmitting a packet in steady state. They assume that in steady state, a station transmits a packet with the probability of p = 1/(E[B] +1) where E[B] is the average value of the backoff timer. Since the average value of the backoff timer can be expressed as E[B] = (E[CW] 1)/2 where E[CW] is the average contention window size of sending a packet, the probability for a packet transmission is obtained by using the average contention window size as p = 2/(E[B] +1) Based on this observation, Ca li et al are able to de rive the following formula for the aggregate network throughput (refer [12]) for detailed derivation procedures): 1 1 1 1 1 1(1)1(1)(1) [[]][] (1) (1) [] [{[(1)(1)]} 1[(1)(1)] (1)MM s M M hM hM s MM hm ppM pp tE c o l l D I F S E s Mp Mpp t Mpp Ecoll hpqpq pMpp q ] (3-1) where m is the average packet length, M is the number of active stations, is the maximum propagation time, q is the parameter for the geometric distribution of packet length, ts is the length of a slot (i.e., aSlotTime), E[coll] is the average collision length, and [](2 ) E smSIFSACKDIFS is the average time to complete a successful packet transmission without any collisions. Now, the aggregate network throughput is derived as a function of the probability of a packet transmission p and the number of active stations M from (3-1), because all other 41

PAGE 42

parameters ( ,,,stmq ) are determined by the simulation configuration. This means that if the number of active stations M is fixed and given, then we can obtain the optimal p value, which maximizes the network throughput. This maximu m throughput is the theo retical throughput limit or analytical upper bound based on th e analysis approach from [12]. In the DTB algorithm, the throughput of the IEEE 802.11 MAC protocol, with an optimal backoff window size tuned to the optimal p value for each M, can be improved significantly. However, the p value, and hence the optimal contention window size, depends on the number of active stations. The DTB method needs to compute the optimal contention window size at runtime based on the estimate of the num ber of active stations. If the estimation is not accurate, the wasted slots and packet collisions will be sign ificant. However, to accurately estimate the number of active stations at run-time is not an easy task for practical wireless local area networks with a distributed conten tion-based MAC protocol. Fast Collision Resolution The major deficiency of the IEEE 802.11 MAC protocol comes from the slow collision resolution as the number of active stations increas es. An active station can be in two modes at each contention period, namely, the transmitti ng mode when it wins a contention and the deferring mode when it loses a contention. When a station transmits a packet, the outcome is either one of the two cases: a successful packet transmission or a collision. Therefore, a station will be in one of the followi ng three states at each contenti on period: a successful packet transmission state, a collision state, and a defe rring state. In most di stributed contention-based MAC algorithms, there is no change in the cont ention window size for the deferring stations, and the backoff timer will decrease by one slot whenever an idle slot is detected. In the proposed fast collision resolution (FCR) algorithm, we will change the contention window size for the 42

PAGE 43

deferring stations and regenerate the backoff timers for all potential transmitting stations to actively avoid future potential collisions. In this way, we can resolve possible packet collisions quickly. More importantly, the proposed algorithm preserves the simplicity of implementation like the IEEE 802.11 MAC. The FCR algorit hm has the following characteristics: Use much smaller initial (min imum) contention window size CWmin than the IEEE 802.11 MAC Use much larger maximu m contention window size CWmax than the IEEE 802.11 MAC Increase the contention window size of a station when it is in both collision state and deferring state Reduce the backoff timers exponentially fast when a prefixed number of consecutive idle slots are detected. Assign the maximum successive packet tran smission limit to achieve good fairness performance. Using these characteristics, in Fast Collisi on Resolution (FCR) [31], the potential sender increases its contention window before a collision actually happens to avoid future collision. To reduce the average number of idle slots, FCR d ecreases the backoff counter exponentially fast when a fixed number of consecutive idle slots were detected. This FCR algorithm indeed resolves collisions faster and re duces the idle slots more effec tively, but could not support good fairness and QoS. Exponentially Increase Exponentially Decrease In EIED [32], whenever a packet transmitted from a node is involved in a collision, the contention window size for the node is increase by backoff factor and the contention window for the node is decreased by backoff factor Ir D r if the node transmits a packet successfully. The performance of EIED is affected by the choice of the values of andIr D r Figure 3-2 show the backoff mechanis m of this EIED algorithm with and 2Ir 2Dr 43

PAGE 44

Table 3-1 summarizes typical backoff algorithms. They all concentrate on adjusting the increasing and decreasing factors of contenti on window size. In gene ral, these algorithms decrease the contention window at the successful transmitter and increase that at the collided transmitter. An important issue is to determ ine how fast these changes should be and how neighboring nodes should respond to th e channel activities [5]. From the table we can see that these algorithms intend to resolve collisions w ith extended window ranges. The larger the new range, the smaller the collision probability. However, they all ignored an im portant fact that the redistributed stations may still locate in or iginal ranges to produ ce further collisions. Table 3-1 Typical backo ff algorithms summarization Algorithm Upon Collision Upon Success Remarks BEB [33] CW = min(2CW, CWmax) CW = CWmin Immediate window resetting results in fairness problem MILD [34] CW = min(1.5CW, CWmax) CW = max(CW-1, CWmin) Window copy scheme introduced window migration problem FCR [31] CW = min(2CW, CWmax) CW = CWmin Deferring nodes double CW on busy state; Nodes reduce backoff counter exponentially fast MIMD [36] CW = min(2CW, CWmax) CW = max(0.5CW, CWmin) Multiplicative increase and multiplicative decrease EIED [32] CW = min( rICW, CWmax), rI > 1 CW = max(CW/ rD, CWmin), rD >1 Exponential increase exponential decrease; If rI > rD, EIED achieves higher performance gain MIMLD [37] CW = min{CWmax, 2max(CW, CWbasic)} if CW > CWbasic CW = max(0.5CW, CWbasic) if CW CWbasic CW = max(CW-1, CWmin) Threshold CWbasic is used to distinguish contention intensity; CWmin CWbasic CWmax GDCF [38] CW = min(2CW, CWmax) After c consecutive successful transmissions: CW = max(0.5CW, CWmin) Flexible for supporting priority access by selecting different values of c for different traffic types 44

PAGE 45

Figure 3-2 Backoff mechanism of EIED: CWmin = 16, CWmax = 1024 45

PAGE 46

CHAPTER 4 CONTENTION WINDOW BASED PARAMETER SELECTION TO IMPROVE POWER LINE MAC EFFICIENCY FOR LARGE NUMBER OF USERS Introduction With the emergence of affordable and portable devices for home, entertainment and personal use, it is natural that the next step would be to get them to communicate with each other on a multimedia and digital platform in the local area network environment of a home or small office/home office. Affordable broadband Intern et communication to residential customers is now available via cable modem and various forms of digital subscriber lines (DSL). While it is a simple matter to use a 10/100 Base-T network hub to link several computers in a single room or in a small office environment, it is much more challenging to provide network connections in several rooms in a typical home [39, 40]. Anothe r approach is to deploy an IEEE 802.11 wireless local area network (WLAN) with wireless mode ms in each device connecting to one or more wireless hubs (infrastructure-base d) or to each other (ad-hoc based). The wireless option is certainly viable (with data rates now up to 54 Mb ps), except for the fact that a dedicated wired infrastructure connecting multiple access points is required to cover the entire home. Power lines, being ubiquitously deployed as a wire-line network for carrying electrical power, are then the obvious choice as the medium for communica tion amongst the plethora of home-based and personal devices like PDAs and MP3 players. They offer the convenience of already being there, and having outlets in almost all locations in a household for easy access. Further, devices can easily obtain electric power if th ey are deployed on PLC systems, while wireless mobile devices rely on batteries and have difficulty in mainta ining continuous electric power. HomePlug 1.0, standardized in 2001, is one of the most popul ar power line communicati on technologies, and it supports up to 14 Mbps transmission rate using power lines. 46

PAGE 47

47 The PLC systems, however, are not free of problems. The power line communication channel can be notorious due to electric noise and interference, as well as channel variability depending on the appliances that are in use from time to time. Even then, with the support of advanced modulation and channel coding tech niques, e.g. Orthogonal Frequency Division Multiplexing (OFDM) and Forward Error Correction (FEC), the present version of the HomePlug 1.0 has been shown to out-perform the traditional IEEE 802.11a/b in many field tests of connectivity, throughput and link stability. The MAC of HomePlug 1.0 is CSMA/CA based and extends the random backoff algorithm of IEEE 802.11. Analytical performan ce evaluation and enhancements of 802.11 MAC has been extensively carried out in [8, 13, 25, 41, 42]. Based on that body of work, Jung et al [43] worked out a detailed analysis for the MAC performance of HomePlug 1.0. In this chapter, a modification in the medium access control protoc ol of Home Plug 1.0 is proposed, to make it a constant contention window based scheme. This modification is shown to significantly enhance the MAC performance under satura tion conditions. An analytical and simulation framework is used to tune the modified protocol for best pe rformance, under the assump tion that the number of active stations is known. The rest of the chapter is organized as follows: section 3.2.1 presents the modification of the standard protocol that makes it a constant contention window scheme, and its analytical model. In section 3.2.2, th e resulting theoretical bound of the MAC efficiency is evaluated, and the constant contention window si ze is parameterized. Section 3.2.3 gives numerical results, and section 3.3 gives some concluding remarks.

PAGE 48

48 Figure 4-1 State space representation of Markov chain model for constant contention window based HomePlug 1.0

PAGE 49

Constant Contention Window MAC The standard HomePlug 1.0 back off algorithm is tailored for networks with small number of stations. In networks with large number of stations, such as may be encountered in BPL applications, the collision overhead drastically reduces the MAC efficiency. To overcome this problem, we propose a new back off scheme that uses and adapts the contention window size based on the number of contending stations in the network. Furthe rmore, the BPC value is kept constant under all circumstances. So, rather than changing the CW every ti me BPC is updated at the end of a collision or success or when DC reac hes a negative value, th e station reverts to the same value of CW (hence called W) to draw the next BC from. Since the value of DC is also updated with BPC, it is reset to the same value, hence called every time BPC is reset. Mathematical Analysis A homogeneous, bi-dimensional Markov chain is used to model the above described system. Each state comprises the current values of DC and BC for the node under consideration. The Markov Chain is in discrete time, in that th e state of the system is observed at the beginning of every slot. Since collisions and successful tran smissions also begin at the start of a slot, the time spent in these, which is more than an actua l slot time, can be embedded into the time of Markov Chain. Figure 4-1 shows the state transm ission diagram of the Markov Chain for a node in the modified HomePlug 1.0. Here, pi represents the probability th at a node finds the medium idle at current time slot. The analysis is carried out for n stations of equal priority under saturation conditions [44]. (i,j) is used to represent the steady state pr obability that the node has a DC value of i and BC value of j For the given Markov Chain, the stati onary distribution can be found from the following two recursive equations: 49

PAGE 50

1 1(,)(1)(1, ) for1,2,...,0,1,...1p p pi j ii jWippWij iWa n d (4-1) and 1 0(,)(,1) for1,2,...i j i jWiWp iW (4-2) The probability p0 that a node will transmit in any randomly chosen slot of time is given by (4-3) 0 0(,0)ip iSince pi is the probability that the node under consideration fi nds the medium idle, it is related to p0 through the following equation: 1 0(1)n ipp (4-4) Solving these equations nu merically, the value of p0 corresponding to n W and can be found. This can be used to calculate the MA C efficiency for the modified HomePlug 1.0. Specifically, the probabilities that a randomly ch osen time slot will experience, respectively, a successful transmission, an idle passage and a collision are: 1 00 0(1) (1) 1n S n I CSIPnpp Pp PPP (4-5) The MAC efficiency, can then be expressed in terms of PS, PI, PC and slot time TI, time for successful transmission TS (inclusive of data time TData response time and RIFS), and time for a collision TC. If the collision time and successful transm ission time are considered constant, and if the time for priority resolution is not considered, the relation becomes 50

PAGE 51

SData SSCCIIPT PTPTPT (4-6) It is shown that the MAC e fficiency found from this math ematical analysis matches exactly that from simulation of the consta nt contention window based HomePlug 1.0. 10 20 30 40 50 60 70 80 90 100 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Number of NodesMAC Efficiency simulation analytical W=16,DC=7 W=16,DC=7 W=16,DC=7 W=16,DC=7 W=32, DC=7 Figure 4-2 Comparison of analytical and simula tion based performance of constant contention window based HomePlug 1.0 Contention Window Parameterization Note that the expression for MAC efficiency in Eq. 4-6 is a function of p0. To maximize the MAC efficiency, this expression can be differentiated with respect to p0 to get the optimal value p0, called p0,opt. For n stations, we get 0000()(SCS I ISICSCdPdPdP dP TPPTPP dpdpdpdp ) 0 (4-7) 51

PAGE 52

0, 0,1 1 (1)opt I n Conp T Tpp t (4-8) It is now possible, using Eq 4-1, 4-2, 4-3, 4-4 and p0,opt found in Eq. 4-8, to parameterize W and to values that will give the best performance for the modified MAC running for a given number of stations. In the next section, the numerical results to these effects are presented. Numerical Results A discrete event simulation was written for generating results for the modified HomePlug 1.0 MAC, and to compare them with those of the standard protocol. Figure 4-2 shows the agreement between the analytical and simulation results for the MAC efficiency of the constant contention window based Ho mePlug 1.0 protocol. Results are s hown for two different cases of values of W, is fixed at 7. Notice that for a constant the performance gets better as W increases. Figure 4-3 shows the solution of Eq. 4-8 for varying number of stations. Many cases of TI and TC are shown. Note that Eq. 4-8 is agnostic to th e underlying backoff scheme. The result shown would be the same for 802.11 and standa rd HomePlug 1.0 for the same values of TI and TC. The rest of the analysis is carried out for values of TI and TC equaling 20 micro-seconds and 800 micro-seconds respectively (i.e. collisions last for 40 slots). Using Eq. 4-1, 4-2, 4-3, and 4-4, the values of p0 can be obtained as a function of W and for a given number of stations. Figu re 4-4 shows the results for n = 5. For Figure 4-4, W is varied from 30 to 50, and from 3 to 15. It is possible to search through this mesh for those values of W and that give p0,opt for that number of stations. Note that there can be more than one combination of W and that give p0,opt. E.g., for n = 5, p0,opt is 0.0446. This corresponds to W = 34 and = 3, and W = 44 and = 15. 52

PAGE 53

0 20 40 60 80 100 120 140 160 180 200 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 Number of nodesOptimal p0 value Optimal p0 for different slot time/collision time ratios Figure 4-3 Optimal p0 value as a function of number of stations; the different curves correspond to varying TI/ TC ratios 0 5 10 15 30 35 40 45 50 0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 Contention Window (W) Defer Counter p0 N = 5 Figure 4-4 The p0 as a function of W and for 5 stations 53

PAGE 54

Performing a search for W and as in Figure 4-4, for a varyi ng number of stations, yields Figure 4-5. Results are shown for = 3 and = 15. In both cases, the W value that gives the optimal p0 is a linear function of the nu mber of stations. In fact, th e slope of the best linear fit doesnt change with only the intercept does. 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 Optimal Contention Window 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 Number of NodesOptimal Contention window DC = 3 DC = 15 W = 5n + 10 W = 5n + 35 Figure 4-5 Variation of W that gives optimal p0 as a function of number of stations, for two values of Using results of figure 4-5, W can be parameterized as a function of number of active stations. Note that both cases in Figu re 4-5 will produce the same value of p0,opt for a given number of stations. Figure 4-6 reports the e nhancement in MAC efficiency when optimal constant contention window, as a function of n, is deployed in the modified HomePlug MAC. 54

PAGE 55

10 20 30 40 50 60 70 80 90 100 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Number of NodesMAC Efficiency original HP 1.0 optimal constant CW based HP Figure 4-6 Comparison of performance of standa rd HomePlug 1.0 MAC a nd constant CW based MAC with optimal window sizes Table 4-1 Pseudo code for the MAC efficiency simulation of HomePlug1.0 and proposed scheme Original HomePlug 1.0 Proposed scheme 4 stages of BPC {at successful transmission, node's BPC = 1st stage of BPC; node's DC = 1st stage of DC; node's BC = randomly select (0, CW0); others' DC = DC 1; others' BC = BC 1; at collision, node's BPC = next stage of BPC; node's DC = new DC; node's BC = randomly select (0, newCW-1); others' DC = DC 1; others' BC = BC 1; if (others' DC < 0) their BPC = next stage of BPC; their DC = new DC; their BC = randomly select (0, newCW-1); end } at idle state all node's BC = BC 1; 1 stage of BPC, 1 optimal CW, 1 DC {at successful transmission or collision node's BPC = BPC; node's DC = DC; node's BC = randomly select (0, CW-1); others' DC = DC 1; others' BC = BC 1; if (others' DC < 0) their BPC = BPC; their DC = DC; their BC = randomly select (0, CW-1); end } at idle state all node's BC = BC 1; end 55

PAGE 56

Conclusions In this chapter, a modification to the standard HomePlug 1.0 MAC protocol has been proposed, to make it a constant contention wi ndow based scheme. An analytical framework based on Markov Chains is deve loped for modeling this modifi ed protocol under saturation conditions, and is proven to accurately match the actual performance of the system. It is shown that the performance can be substantially enhanced if the variables of the modified system (the contention window and defer counter) are parame terized in terms of the number of active stations. This parameterization emerges to be linear, which, when implemented, produces superior results. If the number of stations can be estimated at run time, th e modified protocol can be dynamically tuned to provide optimal performance. 56

PAGE 57

CHAPTER 5 PERFORMANCE ENHANCEMENT OF CONSTANT CW BASED MAC PROTOCOL WITH A NODE ESTIMATION SCHEME Introduction In the previous chapter, we have propos ed optimal contention window based MAC access scheme in power line communicatio ns and proved that it keeps hi gh MAC Efficiency even when there are large number of stations in the cha nnel by simulations. However, the most important key factor to use this novel access scheme is th at every station in the channel should know how many nodes are in the channel at any moment. It is impo ssible in current power line communication environment not using any central devices such as Access Point in wireless communications. For this reason, in this chap ter, we proposed the way that each node can independently estimate the actual number of nodes in the channel. The result of optimal contention window based MAC access scheme app lied by this estimation was implemented by computer simulation under considering of some hidden terminal problems in the network, and proved that it still has great performance even when there are large number of users are in the channel. Estimation of Number of Nodes The success of the above outlined strategy de pends on the knowledge of the number of nodes. In other words, only if every station ha s this knowledge can it tu ne its contention window to the optimal value. In HomePlug 1.0, there is no central coordinator node that might be able to disseminate such information to th e rest of the nodes. It is then up to each node to estimate the network population based on its observation of th e channel conditions. An estimation algorithm which can be independently implemented by each node is presented and analyzed in this section, and is shown to track the networ k population as a function of time 57

PAGE 58

Estimation Strategy Due to the physical carrier sensing ability built into HomePlug, each node can observe the proceedings on the channel and use the equations of the Markov system described above to get estimates of the number of active nodes on the ch annel. In particular, Eq. 4-4 can be written, assuming that p0 is small, as: 01(1)i p np (5-1) If a node can, by observation, estimate the value of pi, it can use the above equation to find the current value of n To get the value of pi, a node uses its observations of the number of idle slots it sees within an ob servation window. Recall that pi is the probability th at a node finds a slot idle, i.e. free from the tr affic of other nodes. Then, pi can be estimated as the ratio of number of idle slots to the obser vation window size. If ni represents the observed number of idle slots in the observation window n0, and if ns is the number of successful tr ansmissions of the node in the same window, then: i i onn p ns (5-2) Finally, the estimate, n of the number of nodes can be found as: () 2 1 2si onn W n n1 (5-3) The value of no must be chosen so that a fair num ber of observations can be made, for statistical accuracy. For example, in [45], the value of the current back off timer is used as the observation window. Since the choice of the back off timer is uniform between 0 and the current contention window size, it might take on small values in many cases, leadi ng to erratic estimates of pi. In this scheme, no is chosen to be the sum of the current contention window size and the 58

PAGE 59

number of back off slots that remain to be counted down at the end of the contention window period [46]. Figure 5-1 shows this in the form of an exam ple. A node calculate its new CW and BC at the Count equals to zero, then, it renews hi s BC only when its BC reaches to zero until the Count is greater then its current CW size. Filter Implementation Count 1 2 3 4 5 6 7 8 9 56 57 58 59 60 61 62 63 64 1 2 BC = 0 BC = 0 BC = 0 BC = 0 Recalculate BC = 50 Recalculate BC = 8 Observation Window Recalculate CW and BC With the Eq. for Node estimation CW = 60, BC = 6 Recalculate CW and BC With the Eq. for Node estimation Figure 5-1 Observation window for updating the estimate of population To smooth the output of the estimation algorithm, an ARMA linear filter is implemented as follows: 1 11q tt innn q ti (5-4) Here, previous q values of the estimate n are averaged and weighed with the factor (1) and added to the current value weighed by to give the next value of the filter output. A value of 10 for q and 0.8 for is chosen, from an analysis based on minimization of the estimation error [45]. 59

PAGE 60

Numerical Results for Estimation The results showing the tracking of number of active nodes in a network, with the employment of the filter in the previous equation, are in Figure 5-2. It shows the active population changing in steps with time, and the track ing ability of the algorithm. The value of the estimate at any time is taken from a node chosen randomly from the entire population. It can be seen from Figure 5-2 that the variance of the estim ate is larger for larger number of nodes. This can be explained by the relationship of Eq. 5-1. The solution is plotted in Figure 5-3. The value of W is chosen to be the op timal value corresponding to = 15. It is clear that as the number of nodes increases, the varia tion in the values of pi becomes very small. So a small disturbance in the calculation of pi results in a large variation of n [47]. 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 x 104 0 20 40 60 80 100 120 140 160 180 Total number of slotsNumber of nodesn=30n=30n=30 n=60 n=100 n=10 n=50 Figure 5-2 Simulation results for estimating number of nodes 60

PAGE 61

0 50 100 150 200 250 300 350 400 450 500 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Number of NodesPiVariation of pi with the number of nodes for optimal CW Figure 5-3 Solution of Eq. 5-1 for pi against number of nodes 0 20 40 60 80 100 120 140 160 180 200 0.77 0.78 0.79 0.8 0.81 0.82 0.83 0.84 Number of NodesNormalized MAC Efficiency Estimation Actual Number of Nodes Figure 5-4 Simulation results of MAC efficiency with estimation scheme 61

PAGE 62

Table 5-1 Pseudo code for the estimation algorithm for figure 5-2 At successful transmission set the actual number of node actual number of node = [30 60 100 10 50; 1 7000 13000 25000 37500]; Inital number of node = 5; {at collion or successful state; observation window size = CWcurrent; count total num ber of idle slot; count total number of successful transmission; stores those values; n(t+1) = alpha*n(t)+((1-alpha )/q*sigma(n(t-i),i=1 to q)); newCW = 5*n(t+1)+15; newBC = randomly(0, CWnew-1); for other nodes; node's DC = DC 1; node's BC = BC 1; busy count++; if (DC < 0) node's DC = initial DC; node's BC = random(0, CWcurrent); end } {at idle state; node's BC = BC 1; } Moreover, since the Markov anal ysis is built assuming that all nodes have the same value of p0, this requires that every node has the same es timate of the current population. However, the proposed algorithm is distributed; each node is implementing it without a ny interaction with the other nodes. No node broadcasts its estimate to the others at any time. As a result of this, there might be differences in the current value of node population held by different nodes. This will cause the nodes to have different values of p0. Even then, the algorith m works reasonably well and the MAC efficiency is not impacted as shown in Figure 5-4. The red line with a star is the optimal MAC Efficiencies of a channel and the blue solid line is the one with the estimation algorithm. As we can see from the Figure 5-4, even the MAC Efficiency with the estimation scheme decreases a little bit as the total number of nodes increases, it is not remarkably low 62

PAGE 63

compared to the optimal values. This also could be explained with the property of pi values mentioned above. Finally, through the simulation of the value of p0 held by all nodes in a network at a particular instant of time, we realized that ther e is not much variation in the p0 values of different nodes, and therefore Eq. 5-1 st ill holds, as an approximation. This proves that there is no need to transmit the popul ation estimate through the network. Conclusions In the previous chapter, we proposed optimal contention window based MAC access scheme in power line communi cation networks with seve ral assumptions. Among those assumptions, the heaviest weighted is Each station knows the actual number of nodes in the same channel. However, it is impossible for each station to know how ma ny stations are in the channel because there is no control device in power line networks like Access Point in wireless network. Another big assumption is There are no hidden terminals in the network. The hidden terminal problems are inevitable in both wireless and power line networks. So, in this chapter, a node estimation scheme which enables to use the optimal constant contention window based MAC access scheme in power line communications has been proposed and the simulation results are shown that this estimation algorithm well ke eps tracking the actual number of nodes in a network even though the estimation variations are little bit high for the large number of nodes. 63

PAGE 64

CHAPTER 6 PERFORMANCE EVALUATION OF MODIFIED HOMPLUG 1.0 MAC PROTOCOL UNDER HIDDEN NODE CONDITIONS Introduction Recently, with the rapid growth of digital appliances and communication technology, the home networking based on Power Line Communication (PLC) netw orks has brought big interest in academic and industry. Power lines, being ubi quitously deployed as a wire-line network for carrying electrical power, offer the convenience of already being there, and having outlets in almost all locations in a household for easy acce ss. Further, devices can easily obtain electric power if they are deployed on PLC systems, while wireless mobile devices rely on batteries and have difficulty in maintaining continuous electr ic power. However, at the early stage of PLC technology, the network performan ce was not enough to satisfy high data rate services such as VoIP and HDTV. The HomePlug Powerline Alliance has standardized HomePlug 1.0 [1], one of the most famous power line communication technologies and it supports up to 14 Mbps transmission rate using power lines. However, the HomePlug 1.0 protocol is not suitable for high data rate services due to its limitation of network throughput and in sufficient QoS. To support high data rate services, the HomePlug AV, the second generati on of PLC technology which supports raw data rate up to 200 Mbps, has been developed and standardized in 2005 [48]. Using the efficient cooperation of high speed PHY and MAC protoc ol, the HomePlug AV protocol can easily support full multimedia home networking thr oughout the whole house including simultaneous High Definition (HD) an d Standard Definition (SD) video distribution, whole-house audio, Voice over IP and high speed Internet in addition to data networking. HomePlug AV and its predecesso r, HomePlug 1.0 use the same CSMA/CA scheme as a MAC layer protocol, which extends the random backoff algorithm of IEEE 802.11 wireless LAN 64

PAGE 65

standard [9]. The existing Home Plug MAC protocol, like IEEE 802.11 protocol, has some severe weaknesses of overall performance. Most of all, it can not provide sufficient MAC efficiency as the number of stations increases in the channel. It is a natural characteristic of contention based protocols where many stations use the sa me channel without pre-coordination. Another critical issue is the hidden node problems in the chan nel. Since PLC channels are in some ways similar to the wireless environm ent, there could be some hidden nodes in the channel. The hidden nodes in the channel increa se the potential collis ion probability in the contention period, which can have very detrimental effect on the overall network performance. To overcome this limitation of contention ba sed MAC protocol, many studies have been proposed under various scenarios in both wire less and power line communications. Analytical performance evaluation and enhancements of 80 2.11 MAC has been extens ively carried out in [8, 13, 25, 41], and Min Young Chung, et al. pr oposed a detailed analys is for the performance evaluation of HomePlug CSMA/CA protocol with constructing a discrete Markov chain model [43]. Based on that body of work, we realized that th ere is close correla tion between the overall MAC performance and contention window size. Fo r example, if we use smaller contention window size than optimal, it may lead to more collisions. Conversely, a bigger one brings about lots of idle states which result in wasting the slot time. With this concept, a modification in the me dium access control protocol of HomePlug 1.0 has been proposed, having an optimal conten tion window size in accordance with the total number of nodes in the network. An analytical and simulation framework is used to tune the modified protocol for the best performance. This requires the knowledge of the active number of users on the network. However, in HomePl ug 1.0 based PLC environment where nodes can communicate each other without a ny access point, it is almost im possible for each node to have 65

PAGE 66

knowledge about the total num ber of nodes in a network. So, in th is chapter, we first assume that every node already knows the total number of node s in a network in order to apply our optimal constant contention window base d MAC protocol scheme to a computer simulation. And then, a node estimation scheme, each node can estimate inde pendently the total number of nodes, also has been proposed to utilize our OCCW algorithm by means of re moving the unrealistic assumption. Finally, it is proven that this modi fied optimal contention window based HomePlug 1.0 protocol sustains high performance under some hidden n ode scenarios in a network by an appropriate computer simulation. The rest of the chapter is organized as follo ws: Section II gives a basic concept of hidden node conditions both in wireless and powerline communication envi ronment. Section III presents the hidden node scenarios, where we ran our si mulation. In section IV the resulting of the simulation based on the conditions, which is built in section 2 is evaluated, and some concluding remarks are presented. Hidden Node Conditions Figure 6-1 Hidden node environment in wireless local area network 66

PAGE 67

Figure 6-2 Hidden node environm ent in power line network Unlike any other wired network protocols such as CSMA/CD, the cost of a collision in the system using collision avoidance is very high be cause a station cannot detect collisions until the entire frame has been transmitted. Moreover, there is a greater vulnerability to collisions due to the presence of hidden terminals in the system. Th ese collisions are severe factor degrading the overall network performance both in wi reless and PLC environment [49-51]. A common solution to the hidde n-node problem is the use of an RTS/CTS handshake before data transmission [52]. The purpose of RTS/CTS is to notify nearby stations of the incipient data transmission period so that those who are not involved in the data exchange will avoid the channel during that pe riod of time. Signal reception of mobile devices in wireless networks usually suffers from nearby activities, r eception is expected to be constantly changing, and thus exchanging RTS/CTS packets before ever y data transmission is required. As shown in Figure 6-1, the access point can communicate wi th both device1 and device2 but those two devices cant communicate each other since they are hidden. If two nodes are hidden each other, they have different view of the channel state, which causes more collisions. 67

PAGE 68

This concept is very similar to the PLC network but channel conditions are more severe in PLC channels. Typical example is asymmetric ch annel conditions due to severe noise. To conquer noisy channels, OFDM modulation is used in the PHY layer protocol. However, the asymmetric channel conditions still remain as a bi g issue to be resolved because it makes hidden node problems more severe. Under asymmetric ch annel conditions, the devices that cannot hear the RTS/CTS can still affect the on going transmi ssions [53]. This is shown in Figure 6-2. Also, in PLC network, once a node is a hidden node from the others in a communication range, it is always a hidden node because of th e nature of a semi-fixed topology. Hidden Node Scenarios As described in the previous section, the hidden node problem in Power Line Communications is also one of the most important factors having an influence on the overall network performance like in WLAN. Though a RTS/CTS scheme has been widely used as a solution to conquer this hidden node problem, it is also just one level solution. Like this, the hidden node problem is a critical issue which can not be passed over both in wireless and power line communications. At the beginning of this chapter, we have implemented our constant contention window based MAC pr otocol under the two unrealistic assumptions that knowledge about the total number of nodes in a network and ideal channel conditions. They were essential to follow up the discrete Markov Chain model and all derived probabilities, which we have used to analyze our modified algorithm mathematically To remove these unrealistic assumptions, we proposed node estimation scheme which every n ode can estimate the total number of nodes in the network, and proved that this proposed scheme fairly well performs the estimation in Chapter 5. Now, we will prove that this proposed al gorithm also has a great performance under various hidden node scenarios even it has b een derived under ideal channel conditions. A 68

PAGE 69

computer simulation has been carried out to see the MAC efficiency and detailed processes are as follows. 0 10 20 30 40 50 60 70 80 90 100 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Number of nodesMAC Efficiency original HP1.0 with 1 pair of hidden proposed scheme with 1 pair of hidden and node information proposed scheme with 1 pair of hidden and estimation A 0 10 20 30 40 50 60 70 80 90 100 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Number of nodesMAC Efficiency original HP1.0 with 2 pairs of hidden proposed scheme with 2 pairs of hidden and node information proposed scheme with 2 pairs of hidden and esitmation B 0 10 20 30 40 50 60 70 80 90 100 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Number of nodesMAC Efficiency original HP1.0 with 3 pairs of hidden proposed scheme with 3 pairs of hidden and node information proposed scheme with 3 pairs of hieedn and estimation C 0 10 20 30 40 50 60 70 80 90 100 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Number of nodesMAC Efficiency original HP1.0 with 4 pairs of hidden proposed scheme with 4 pairs of hidden and node information proposed scheme with 4 pairs of hidden and estimation D 0 10 20 30 40 50 60 70 80 90 100 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Number of nodesMAC Efficiency original HP1.0 with 5 pairs of hidden proposed scheme with 5 pairs of hidden and node information proposed scheme with 5 pairs of hidden and estimation E 0 10 20 30 40 50 60 70 80 90 100 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Number of nodesMAC Efficiency original HP1.0 with 20% hidden pairs proposed scheme with 20% hidden pairs and node information proposed scheme with 20% hidden pairs and estimaion F Figure 6-3 The MAC efficiency under various hidden terminal conditions. 69

PAGE 70

Our simulation considered a network of five to hundred nodes having several pairs of hidden nodes in each case. Simulation parameters ar e exactly same values used in Chapter 5 and we did not take account of RTS/CTS control pa cket while transmitting data. Also we assumed that there is no loss of data p ackets except at collision states. Figure 6-3 shows the simulation result of the overall MAC efficiency under various hidden node conditions. The blue lines with a star represent the pe rformance result of original HomePlug 1.0 MAC protocol under di fferent hidden pairs up to maximum 20 percent, red lines with a circle represen t the performance result of modified Constant Contention Window based MAC protocol under the same porti on of hidden pairs with the a ssumption that all nodes have node information, and the black lines with a square are the pe rformance result of proposed scheme under the same portion of hidden pairs w ith estimation scheme. We assumed that if node i and node j are hidden each other, node i or node j can not be a part of different hidden pair. As we can see from the figure, the overall MAC Efficiencies of proposed algorithm under the assumption that each node already knows the actual number of nodes in the channel are around 0.7 where 20% of hidden nodes are included and in creased up to the optimal values derived by mathematical analysis. However, in case of the original HomePlug 1.0 MAC, the MAC Efficiencies are getting smaller as the total number of nodes are in creased even they contain the same portion of hidden nodes. Although the origin al HomePlug 1.0 MAC has a little bit better performance for small number of nodes under the same hidden node scenarios, it is just because proposed algorithm is based on the Discrete Ma rkov Chain model assuming the ideal channel conditions. The proposed scheme with estima tion scheme also did not provide high MAC efficiency under 20% hidden node conditions. In proposed estimation scheme, since each node estimate total number of nodes in the channel independently, the transmission between the 70

PAGE 71

hidden nodes can be counted as a successful tran smission, which returns wrong estimation result. Consequently, the proposed constant conten tion window based modifi ed HomePlug 1.0 MAC protocol can support outstanding performance even under various hidden terminal conditions if all node have information. Simulation Results and Conclusions In this chapter, a modification of the standa rd HomePlug 1.0 MAC protocol [44] has been simply reminded, to make it a constant c ontention window based scheme. An analytical framework based on Markov Chains has been deve loped for modeling this modified protocol under saturation conditions, and proven to accurately match the actual performance of the system. It has been shown that th e performance can be substantially enhanced if the variables of the modified system (the conten tion window and defer counter) are parameterized in terms of the number of active stations. However, this pr oposed Optimal Contention Window based back off algorithm has been implemented under critical two unrealistic assumptions; every node already knows the total number of stati ons in the network and ideal channel conditions. In PLC environment, it is almost impossible for each nod e to get the node inform ation because there is no central point such as AP in wireless networ k environment. Also, the PLC channel is much harsher condition than wireless due to severe noise problem. To conquer this unrealistic assumptions, a distributed node estimation scheme that each node can estimate the total number of nodes in the network independently, has been proposed and showed that this estimation algorithm performs well tracking the varying num ber of nodes and the overall MAC Efficiency is also very near to the optimal values thr ough an appropriate computer simulation. Finally, a performance evaluation of th is proposed algorithm under hidd en node scenarios has been performed since this is a very important factor in PLC channel due to the nature of a semi-fixed 71

PAGE 72

topology, and it is proved that this modified algorithm also has a reliable performance even it has been derived with discrete Markovian model under the assumption of ideal channel condition. 72

PAGE 73

CHAPTER 7 AN EFFICIENT TCP MECHANISM TO RE DUCE OUT-OF-SEQUENCE PACKETS IN MOBILE IPV6 Introduction In Mobile IPv6, to perform packet transmi ssion continuously without disconnection of the L3 layer, the mobile node has a home IP addres s (HoA) for identification and a temporal IP address for routing information [54 2003]. When an MN moves to ne w subnet, it should disconnect with the current access router and connect with new access router, to obtain a new temporal address called care-of address (CoA). Then MN shoul d register the binding new CoA with its HoA to its home agent (HA) and its CNs. That is, we need to know the movement detection time, new CoA configuration time and binding update time to start internet service from new subnet network. During handover, the transmitting packets from a HA or CN may be lost. Recent work has been aimed at improving Mobile IPv6 handover performance in order to support real time and other delay sensitive traffic. Packet losses during handover are treated as an indication of network congestion, which causes TCP to take unnecessa ry congestion avoiding measures [55]. For this work, some trials have been proposed such as Smooth handover for Mobile IP by Route Optimization in mobile IP [56] and Fast Handover for Mobile IPv6 [57 2005] To reduce the handoff latency and packet loss, Mobile IPv6 [54 2003] develops the fast handoff protocol [57 2005]. By the help of link la yer, the fast handoff de tects change of link connectivity and sets up a routing path, so cal led tunnel, between two access routers (AR). The tunnel is used to forward packets to a new AR from a previous AR, until a mobile anchor point (MAP) changes the routing path from the previous AR to the new AR. However, the Fast Mobile IPv6 (FMIPv6) [57 2005] may cause packet mis-ordering problem between tunneling packets from PAR, HA and directly delivered packets from CN. 73

PAGE 74

These out-of-sequence packets encounter perfor mance decline of TCP by duplication of the ACKs from the TCP congestion control on the transport laye r and induce useless packet retransmissions from CN. In real time service applications, it is difficult to remedy the out-of-sequence packet correctly. Thus, a few trials have been develope d to solve this problem such as Design and Analysis of the mobile agent preventing out of sequence packets [ 58 1999] and performance improvement by packet buffering [59]. In this chapter, we proposes the use of an efficient TCP mechanism in FMIPv6 to prevent a mis-ordering of data packets during handover, speci fically by adding a new reordering scheme to the base station connected to the NAR with a mo dified TCP header form at at a source device such as a HA or CN. Related works and problems Standard mobile IPv6 Figure 7-1 Standard mobile IPv6 handover procedure 74

PAGE 75

Figure 7-1 shows standard Mobile IPv6 message flow for route optimization in Mobile IPv6. In Mobile IPv6 network, an MN could move out of reach in the middle of a packet transfer [54 2003]. The binding update message is used to info rm both HA and CN of MNs new location. After the handover procedure, the MN is going to experience a black-out state and there is no packet exchange. At this time, a number of pack ets would be delivered to the old foreign router (FRold) rather than the new foreign router (FRnew). These packets that lose their route are called old packets. Mobile IPv6 allows the MN to send a binding update message to the FRold in previously visited IP subnet. Then the FRold forwards the old packets sent by the CN to the FRnew using tunneling mechanism so that the MN can receive packets wit hout any loss. This forwarding is known as smooth handover in Mob ile IP. After receivi ng the binding update message, the CN re-routes the packets to the FRnew. That is, when route optimization is used, the mobile node sends its NCoA to the CN using binding update (BU) messages. After receiving BU, the packets leavi ng the CN are routed directly to the NCoA. However, before the CN re ceives the BU, it continues to route packets to the NCoA via HA. Consequently, these two types of packets will be arrived out-of-sequence at the NAR. Figure 7-2 shows an example of the out-of-se quence packet problem in standard Mobile IPv6 network. In figure 7-2, the packets (from 1 to 5), which have already been in-flight until the binding change of the CN, are to be forwarded to the new foreign rout er (FR) by tunneling mechanism. These tunneled packets are out-of-sequence, since packet 6 from CN directly reaches new FR earlier than the tunneled packets (from 1 to 5) from PAR. 75

PAGE 76

Figure 7-2 Out-of-sequence packet pr oblem in a mobile IPv6 network Fast Mobile IPv6 Figure 7-3 shows packet transmission during handover in Fast MIPv6 networks. The Fast Handover Protocol [57 2005] is an extension of Mobile IPv6 that a llows an AR to offer services to an MN in order to anticip ate a layer 3 (L3) handover. Movement anticipation is based on the layer 2 (L 2) triggers. An L2 trigger is information based on the link layer protocol, below the IPv6 pr otocol, in order to begin L3 handover before L2 handover ends. The MN obtains a new CoA (NCo A), while still connected to the PAR, by means of router advertisements, containing network information, from the NAR. The PAR validates the MNs new CoA and initiates the process of establishi ng a bidirectional tunnel between the PAR and the NAR by sending a Handover Initiate (HI) message to the NAR. Then, the NAR verifies that its 76

PAGE 77

new CoA can be used on the NARs link. Also the NAR sets up a host route for the MNs previous CoA (PCoA) and responds with a Handover Acknowledge (HACK) message. Figure 7-3 Fast handover for mobile IPv6 When the MN receives a PrRtAdv message, it should send a Fast Binding Update (F-BU) message, preferably prior to disconnecting its link. When the PAR recei ves an F-BU message, it must verify that the requested handover is accepted by the NAR as indicated in the HACK message status code. Then it begins forwardi ng packets intended for PCoA to the NAR and sends a Fast Binding Acknowledgeme nt (F-BACK) message to the MN. After the change of the link connectivity with the NAR, the MN and NAR exchange a Router Soli citation (RS) Message with Fast neighbor Advertisemen t (FNA) option and a Router Adve rtisement message (RA) with Neighbor Advertisement Acknowledgment (NAAC K) option. After the NAR sends Router Advertisement message with NAACK option, it starts to deliver buffered packets tunneled from PAR and buffered packets from CN directly. Until the CN receive s BU, the packets sent from 77

PAGE 78

CN are tunneled from PAR to NAR. After the CN receives BU, the CN di rectly delivers the packets to an MN. Consequently, if the distance between CN and NAR is shorter than tunneled distance from CN to NAR via PAR, the MN may receive out of sequence packets. Figure 7-4 Out-of-sequence pack et problem in FMIPv6 network Figure 7-4 shows out-of-sequence packet problem in fast Mob ile IPv6 (FMIPv6). After the PAR sends fast binding update message to the NAR and the MN, packet 3 to 5 are tunneled from the PAR to the NAR and these tunneled packets are buffered in the NAR until the NAR receives a RS from the MN. When the CN receives the BU message from the MN for route optimization, the CN sends packet 6 to 7 directly to the NAR Consequently, if the distance between the CN and NAR is shorter than tunneled distance from th e CN to NAR via PAR, buffered packets in the NAR would be out-of-seqeunce p ackets due to the packet delay time occurred by tunneling. 78

PAGE 79

Improving TCP Mechanism to Redu ce Disordered Packet (I-TCP) I-TCP for movement detection In movement detection, the MN is aware of performing handover to another AP because of the channel maintenance or L3 handover. The MN performs a scan to see APs through probes. Our proposed scheme is based on the snoop prot ocol which prevents DACK and controls TCP packet data sequence [60]. The modified access point (MAP) with snoop agent consists of MAP controller, MAP buffer and sequence checker as shown in Figure 7-5. Figure 7-5 Proposed EAPMIPv6 signaling flow I-TCP for data Figure 7-6 shows the flowchart of I-TCP scheme for data to perform proposed scheme. In this dissertation, we assume that the MAP buf fer size is enough to cach e the received packets directly from the CN until finishing the packet tr ansmission from the PAR. Moreover, to prevent packet overflow in the MAP buffer, the NAR send control messages pe riodically for notifying buffer states to the CN or HA. Using this message the CN or MN can control the data traffic. 79

PAGE 80

Figure 7-6 Proposed I-TCP scheme for data Firstly, during handover in Fast Mobile IPv6, when PAR receives F-BU message, the PAR sends HI message with PARs snoop information to control TCP packet sequence. After the PAR sends F-BACK message to both the MAR and MN buffered packets in PAR are delivered to NAR. When the CN gets the BU message, it sends rest of the packets to the NAR directly. Also, the CN sends last packet with modified TCP header to PAR. This modified last packet is called MLP. To distinguish last packet among the all rece ived packets in the NAR the TCP packet can be modified by adding 1 bit MLP flag to the reserved flag in TCP h eader. If the MLP flag is the packet acts as normal packet. However, the ML P flag is this packet act as last packet from PAR. 80

PAGE 81

This MLP flag is important to distinguish between packets delivered by tunneling from PAR and packets delivered directly from the CN without tunneling. Until the NAR gets the MLP with flag bit the packet delivered di rectly from CN is buffered at MAP buffer. Simulation Results We evaluate the performance of the proposed scheme using the Network Simulator (NS-2) [61]. Based on the standard NS-2 distribution ve rsion ns-allinone2.1b6, the simulation code used for the experiments was designed on top of the INRIA/Motorola MIPv6 code for ns-2 implementation. For the purpose of the performance evalua tion of our proposed R-MIPv6 we used a simulation topology for most of the simulation as shown in Figure 7-7. In TCP code, we used TCP-NEWRENO. Figure 7-7 Simulation network scenario 81

PAGE 82

For the simulation network we us ed wired links with bandwidth of 5Mps and link delay of 20ms, 50ms, and 100ms respectively. All mobile nodes in the simulation m ove linearly from one access router to another at a constant speed of 50m /s. Bulk data transfer (by FTP) is connected between the CN and the MN [62]. Figure 7-8 Packet sequence in standard MIPv6 during handover Figure 7-8 shows the packet transmission in standard Mobile IPv6 during handover procedure. In figure 7-8, some of the packets are lost which necessitates packet retransmission in Mobile IPv6. The packet loss o ccurs because the old route is lost before MN can send binding message to HA. That is, if handover delay time is long, packet loss and out-of-sequence packets would increase in the Mobile IPv6 network. Th is problem causes packet re-transmission which leads to TCP performance degradation. 82

PAGE 83

Figure 7-9 Packet sequence in FMIPv6 during handover Figure 7-10 Packet sequence in proposed I-TCP during handover 83

PAGE 84

Figure 7-9 shows the packet transmission asso ciated with buffering in the NAR and PAR during handover. Although the packet loss does no t happen, the received packet sequence is changed in an MN due to an out-of-sequence packet problem caused by packets received by tunneling and packets recei ved directly from CN. In figure 7-10, although packet delay occurr ed, the receiver can receive the packet normally without packet loss and out-of -sequence packets so that retransm ission is not required. Conclusions In this chapter, we presented fast mobile IPv6 with I-TCP scheme that can improve TCP performance. The use of a proposed R-MIPv6 ca n prevent the out-of-sequence packets problem in existing Mobile IPv6 network. The simulatio n results show that the MIPv6 with I-TCP scheme gives a better performance than FMIPv6. Accordingly, the proposed scheme is a strong candidate to support seamless mobility in MIPv6 based mobile networks with out packet mis-order problems. Our future work consists of further analyz ing and simulating the MI Pv6 packet reordering method in UDP traffic environment. The modifi ed packet frame processing in the enhanced access point inflicts extra processing load. Also the buffer state in the access point can cause overflow problems in large Mobile IPv6 network environments. These factors are to be simulated further with ns-2. The simulations wi ll be expanded with the FMIPv6 implementation. 84

PAGE 85

CHAPTER 8 CONCLUSIONS AND FUTURE WORK In this dissertation, we proposed an optim al contention window based MAC access scheme for power line communication network and an e fficient TCP mechanism by using snoop TCP in Mobile IPv6 network. An analyt ical framework based on Markov Ch ain state transition diagram was developed for modeling this modified pr otocol under saturation conditions, and it was proven to accurately match the actual performanc e of the system. Using this modified scheme provided almost constant overall MAC efficien cy of about 80% even under heavy traffic network conditions. However, the proposed modified MAC sche me has been implemented under several assumptions. Among those assumptions, the heavie st weighted was Eac h station knows the actual number of nodes in the same channel. But, it is impossible for each station to know the node information in the channel because there is no control device in PLC network such as Access Point in wireless network. Another big assumption was There are no hidden terminals in the network. But, the hidden terminal problems are also inevitable in both wireless and power line networks. So, in this di ssertation, a node estimation scheme in which each node can independently estimate the total number of statio ns in the channel at run time has been proposed and shown to produce the required network performance even with its approximation. Also, the throughput simulation of proposed MAC access scheme under various hidden pairs within the network has been proposed and proved that this modified algorithm also had a reliable performance even it had been derived with disc rete Markovian model un der the assumption of ideal channel condition. Now, the broadband communication over Power Line Networks has attracted much interest in academe and industry. In a wide area PLC network, transmitting a packet from a 85

PAGE 86

source to a long distance destina tion, not immediately reachable node requires the packet relay of the intermediate nodes (repeaters). The introduction of repeaters in the ne twork has the potential to increase the global capacity of the PLC system but much more signal interferences should be considered. Also, considering the dynamic topology change and impossible prediction of the power line attenuation, repeater it self can not be statically conf igured. Therefore, the extended researches on the various hidden node problems taking place in a wide range network and on designing an efficient routing protocol for dynamically adapti ng the power line circumstances should be needed. For TCP mechanism in mobile IPv6, a snoop pr otocol which prevents DACK and controls TCP packet data sequence has been applied. Th e use of this proposed R-MIPv6 can prevent the out-of-sequence packets problem which may happen during the tunneling procedures in existing Mobile IPv6 network. A computer simulation of this proposed I-TCP scheme has been performed and the results showed that MIPv6 with I-TCP scheme gave a better performance than FMIPv6. However, the modified packet frame processi ng in the enhanced access point inflicts extra processing load. Also the buffer state in the access point can cause overf low problems in large Mobile IPv6 network environments Further researches to figure these problems out should be required. 86

PAGE 87

APPENDIX A PROGRAMMING SOURCE CODES Original Homeplug 1.0 MAC Backoff Algorithm clear all; close all; NumOfNodes = [5:5:50,60:10:100]; CWs = [8 16 16 32]; % Contention Windows BPC = [0 1 2 3]; % BackOff Procedure Counter DC = [0 1 3 15]; % Deferral Counter T_NumOfSim = 100000; % Total Number of Simulation BPC_Cmn = 1; DC_Cmn = 2; BC_Cmn = 3; T_Col_Cnt = zeros(1,max(size(NumOfNodes))); % Initialize Total Number of Collisions T_Suc_Cnt = zeros(1,max(size(NumOfNodes))); % Initialize Total Number of Success T_Idle_Cnt = zeros(1,max(size(NumOfNodes ))); % Initialize Total Number of Idles Tslot = 20; % Slot Time TSIFS = 0;%.5*Tslot; % SIFS Time TDIFS = 0;%2.5*Tslot; % DIFS Time TACK = 0;%5*Tslot; % ACK Time Tsuc_vec = 40*Tslot; % Succe ssful Packet Transmission Time Tcol_vec = Tsuc_vec; % Co lliding Packet Transmission Time for Node_index = 1:max(size (NumOfNodes)), % How ma ny nodes in the simulation NumNodes = NumOfNodes(Node_index); Nodes = zeros(NumNodes,3); for NodeNum = 1:NumNodes %%%%%% INITIAL STATE %%%%%% Nodes(NodeNum,BPC_Cmn) = BPC(1); Nodes(NodeNum,DC_Cmn) = DC(1); Nodes(NodeNum,BC_C mn) = floor((CWs(1)-1)*rand(1,1)); end for NumOfSim = 1:T_NumOfSim, ChkNodes = find(Nodes(:,BC_Cmn) == 0); [ChkCol temp] = size(ChkNodes); if (ChkCol > 1) %%%%%%%%%%%% ('COLLI SION') %%%%%%%%%%%% T_Col_Cnt(Node _index) = T_Col_Cnt(Node_index) + 1; for Node_element = 1:NumNodes, if (Nodes(Node_element,BC_Cmn) == 0) Nodes(Node_e lement,BPC_Cmn) = Nodes(Node_element,BPC_Cmn) + 1; 87

PAGE 88

if (Nodes(Node_element,BPC_Cmn)) > BPC(3) Nodes(Node_element,BPC_Cmn) = BPC(4); end Nodes(Node_ele ment,DC_Cmn) = DC(Nodes(N ode_element,BPC_Cmn) + 1); Nodes(Node_ele ment,BC_Cmn) = fix(CWs(Node s(Node_element,BPC_Cmn) + 1)*rand(1)); else Nodes(Node_e lement,DC_Cmn) = Nodes(Node_element,DC_Cmn) 1; Nodes(Node_e lement,BC_Cmn) = Nodes(Node_element,BC_Cmn) 1; if (Nodes(Node_element,DC_Cmn) < 0) Nodes(N ode_element,BPC_Cmn) = Nodes(Node_element,BPC_Cmn) + 1; if (Nodes(Node_element,BPC_Cmn)) > BPC(3) Nodes(Node_element,BPC_Cmn) = BPC(4); end Nodes(Node _element,BC_Cmn) = fix(CWs(N odes(Node_element,BPC_Cmn) + 1)*rand(1)); end end end elseif (ChkCol == 1) %%%%%%%%%%%('S UCCESS')%%%%%%%%%%% T_Suc_Cnt(Node _index) = T_Suc_Cnt( Node_index) + 1; for Node_element = 1:NumNodes, if (Nodes(Node_element,BC_Cmn) == 0) Nodes(Node_element,BPC_Cmn) = BPC(1); Nodes(Node_ele ment,DC_Cmn) = DC(Nodes(N ode_element,BPC_Cmn) + 1); Nodes(Node_ele ment,BC_Cmn) = fix(CWs(Node s(Node_element,BPC_Cmn) + 1)*rand(1)); else Nodes(Node_e lement,DC_Cmn) = Nodes(Node_element,DC_Cmn) 1; Nodes(Node_e lement,BC_Cmn) = Nodes(Node_element,BC_Cmn) 1; if (Nodes(Node_element,DC_Cmn) < 0) Nodes(N ode_element,BPC_Cmn) = Nodes(Node_element,BPC_Cmn) + 1; if (Nodes(Node_element,BPC_Cmn)) > BPC(3) Nodes(Node_element,BPC_Cmn) = BPC(4); end Nodes(Node _element,BC_Cmn) = fix(CWs(N odes(Node_element,BPC_Cmn) + 1)*rand(1)); end end end else %%% %%%%%%%%%%('IDLE' )%%%%%%%%%%%%%% T_Idle_Cnt(Node_ index) = T_Idle_Cnt (Node_index) + 1; Nodes(:,BC_Cmn) = Nodes(:,BC_Cmn) 1; 88

PAGE 89

end end MAC_Eff(Node_index) = (T_S uc_Cnt(Node_index)*Tsuc_vec)/... (T_Suc_Cnt(Node_inde x)*(Tsuc_vec+TACK+TSIFS+TDIFS) +... T_Col_Cnt(Node_index)*(Tcol_v ec+TDIFS) + T_Idle_Cnt(Node_index)*Tslot); %MAC_Eff(Node_index) = 40*T_Suc _Cnt(Node_index)/Tslot*T_NumOfSim; end plot(NumOfNodes, MAC_Eff,'r-');gr id on;axis([1 100 0 1]); hold on; The IEEE 802.11 Backoff Algorithm clear all; close all; NumOfNodes = [5:5:20,30:10:100]; CW_min = 32; CW_max = 256; T_NumOfSim = 50000; T_Col_Cnt = zeros(1,max(size(NumOfNodes))); T_Suc_Cnt = zeros(1,max(size(NumOfNodes))); T_Idle_Cnt = zeros(1,max(size(NumOfNodes))); Tslot = 20; TSIFS = .5*Tslot; TDIFS = 2.5*Tslot; TACT = 3*Tslot; Tsuc_vec = 165*Tslot; Tcol_vec = Tsuc_vec; Chk_Fair = zeros(max(size(NumO fNodes)),max(NumOfNodes)); for Node_index = 1:max( size(NumOfNodes)), NumNodes = NumOfNodes(Node_index); Nodes = 0; %% %%%% INITIAL STATE %%%%%% CW = CW_min ones(1,NumNodes); Nodes = floor((CW_min-1)*rand(1,NumNodes)); Suc_Cnt_Each = zeros(1,NumNodes); for NumOfSim = 1:T_NumOfSim, ChkNodes = find(Nodes == 0); [temp ChkCol] = size(ChkNodes); if (ChkCol > 1) T_Col_Cnt(Node _index) = T_Col_Cnt(Node_index) + 1; for Node_element = 1:NumNodes, if (Nodes(Node_element) == 0) previouscw = CW(Node_element); 89

PAGE 90

Nodes(Node _element) = fix((CW(Node_e lement)-1) rand(1)); end end elseif (ChkCol == 1) %fprintf('SUCCESS'); T_Suc_Cnt(Node _index) = T_Suc_Cnt( Node_index) + 1; for Node_element = 1:NumNodes, if (Nodes(Node_element) == 0) Suc_Cnt_Each(Node_element) = Suc_Cnt_Each(Node_element)+1; Nodes(Node _element) = fix((CW(Node_e lement)-1) rand(1)); end end else T_Idle_Cnt(Node_ index) = T_Idle_Cnt (Node_index) + 1; Nodes = Nodes 1; end end MAC_Eff(Node_index) = (T_S uc_Cnt(Node_index)*Tsuc_vec)/... (T_Suc_Cnt(Node_inde x)*(Tsuc_vec+TACT+TSIFS+TDIFS) +... T_Col_Cnt(Node_index)*(Tcol_v ec+TDIFS) + T_Idle_Cnt(Node_index)*Tslot); end plot(NumOfNodes, MAC_Eff,'b'); xlabel('# of Nodes'); ylabel('MAC_Efficiency'); grid on;axis([0 100 0 1]);hold on; Optimal Constant Contention Window based MAC clear all; closed all; NumOfNodes = [5 :5:50 60:10:100]; BPC = 1; % BackOff Procedure Counter DC = 3; % Deferral Counter T_NumOfSim = 20000; % Total Number of Simulation BPC_Cmn = 1; DC_Cmn = 2; BC_Cmn = 3; T_Col_Cnt = zeros(1,max(size(NumOfNodes))); % Initialize Total Number of Collisions T_Suc_Cnt = zeros(1,max(size(NumOfNodes))); % Initialize Total Number of Success T_Idle_Cnt = zeros(1,max(size(NumOfNodes ))); % Initialize Total Number of Idles Tslot = 20; % Slot Time for Node_index = 1:max(size (NumOfNodes)), % How ma ny nodes in the simulation 90

PAGE 91

NumNodes = NumOfNodes(Node_index); Nodes = zeros(NumNodes,3); CWs = 5*NumNodes + 10; for NodeNum = 1:NumNodes %%%%%% INITIAL STATE %%%%%% Nodes(NodeNum,BPC_Cmn) = BPC; Nodes(NodeNum,BC_Cmn) = floor((CWs-1)*rand(1,1)); end for NumOfSim = 1:T_NumOfSim, ChkNodes = find(Nodes(:,BC_Cmn) == 0); [ChkCol temp] = size(ChkNodes); if (ChkCol > 1) %%%%%%%%%%%% ('COLLI SION') %%%%%%%%%%%% T_Col_Cnt(Node _index) = T_Col_Cnt(Node_index) + 1; for Node_element = 1:NumNodes, if (Nodes(Node_element,BC_Cmn) == 0) Nodes(Node_element,BC_Cmn) = fix(CWs*rand(1)); else Nodes(Node_e lement,DC_Cmn) = Nodes(Node_element,DC_Cmn) 1; Nodes(Node_e lement,BC_Cmn) = Nodes(Node_element,BC_Cmn) 1; if (Nodes(Node_element,DC_Cmn) < 0) Nodes(Node_element,DC_Cmn) = DC; Nodes(Node_element,BC_Cmn) = fix(CWs*rand(1)); end end end elseif (ChkCol == 1) %%%%%%%%%%%('S UCCESS')%%%%%%%%%%% T_Suc_Cnt(Node _index) = T_Suc_Cnt( Node_index) + 1; for Node_element = 1:NumNodes, if (Nodes(Node_element,BC_Cmn) == 0) Nodes(Node_element,BC_Cmn) = fix(CWs*rand(1)); else Nodes(Node_e lement,DC_Cmn) = Nodes(Node_element,DC_Cmn) 1; Nodes(Node_e lement,BC_Cmn) = Nodes(Node_element,BC_Cmn) 1; if (Nodes(Node_element,DC_Cmn) < 0) Nodes(Node_element,DC_Cmn) = DC; Nodes(Node_element,BC_Cmn) = fix(CWs*rand(1)); end end end else %%% %%%%%%%%%%('IDLE' )%%%%%%%%%%%%%% T_Idle_Cnt(Node_ index) = T_Idle_Cnt (Node_index) + 1; 91

PAGE 92

Nodes(:,BC_Cmn) = Nodes(:,BC_Cmn) 1; end end MAC_Eff(Node_index) = (T_S uc_Cnt(Node_index)*Tsuc_vec)/... (T_Suc_Cnt(Node_inde x)*(Tsuc_vec+TACK+TSIFS+TDIFS) +... T_Col_Cnt(Node_index)*(Tcol_v ec+TDIFS) + T_Idle_Cnt(Node_index)*Tslot); %MAC_Eff(Node_index) = 40*T_Suc _Cnt(Node_index)/Tslot*T_NumOfSim; end plot(NumOfNodes, MAC_Eff,'r*-');grid on;axis([1 100 0.3 1]); hold on; Estimation Algorithm for Tracki ng the Actual Number of Nodes clear all; close all; clc; Node_Samples = [30 60 100 10 50;1 7000 13000 25000 37500]; NumOfNodes = Node_Samples (1,1);%[5:5:50,60:20:200]; % Actual Number of Nodes DC = 3; % Deferral Counter T_NumOfSim = 50000; % Total Number of Simulation BusyCount = 0; AvgNumOfDCrst = 0; n_init = 5; %NumOfNodes;%StdTa ble(Table_Index_Node,Current); CW_Cmn = 1; DC_Cmn = 2; BC_Cmn = 3; Busy_Cmn = 4; SUC_Cnt = 5; N_Est_Cmn = 6; Coll_Cmn = 7; T_Col_Cnt = 0;%zeros(1,max(size(NumOfNodes))); % Initialize Total Number of Collisions T_Suc_Cnt = 0;%zeros(1,max(size(NumOfNodes)) ); % Initialize Tota l Number of Success T_Idle_Cnt = 0;%zeros(1,max(size(NumOfNodes))); % Initialize Total Number of Idles al = 0.8; Each_N_History = 10; Each_Count = zeros(Each_N_H istory+3,NumOfNodes); Each_Count(3,:) = n_init; Tslot = 20; % Slot Time TSIFS = 0;%.5*Tslot; % SIFS Time TDIFS = 0;%2.5*Tslot; % DIFS Time TACK = 0;%5*Tslot; % ACK Time Tsuc_vec = 40*Tslot; % Succe ssful Packet Transmission Time 92

PAGE 93

Tcol_vec = Tsuc_vec; % Co lliding Packet Transmission Time Count = 0; Nodes = zeros(NumOfNodes,7); Nodes(:,N_Est_Cmn) = n_init; Nodes(:,CW_Cmn) = 5 n_init + 15; %%%%%% Initia lize CW sizes %%%%%% Nodes(:,DC_Cmn) = DC; %%%%%% Initiali ze DC values %%%%%% Nodes(:,SUC_Cnt) = 0; %%%%%% Init ialize Success Count for check fairness %%%%%% Nodes(:,Busy_Cmn) = 0; %% %%%% Initialize Total # of DC Resets %%%%%% Nodes(:,Coll_Cmn) = 0; for NodeNum = 1:NumOfNodes % Initialize BackOff Conut for each nodes % Nodes(NodeNum,BC_Cmn) = fix( Nodes(NodeNum,CW_Cmn)*rand(1,1)); end for NumOfSim = 1:T_NumOfSim, %%%%%% Total # of Simulation %%%%%%% switch NumOfSim case Node_Samples(2,1) NewNumOfNodes = Node_Samples(1,1); if (NewNumOfNodes > NumOfNodes) diff_size = NewNumOfNodes NumOfNodes; temp_Nodes = zeros (diff_size,7); temp_Nodes(:,N_Est_Cmn) = n_init; temp_Nodes(:,CW_Cmn) = 5 n_init + 15; %%% Initialize CW sizes %%% for Node Num = 1:diff_size %% Initialize BackOff Conut for each nodes %% temp_Nodes(NodeNum,BC_Cmn) = fix(temp_Nodes(NodeNum, CW_Cmn)*rand(1,1)); end Each_Count = [Each_Count zeros(Each_N_H istory+3,diff_size)]; NumOfNodes = NewNumOfNodes; Nodes = [Nodes;temp_Nodes]; else Each_C ount = [Each_Count(:, 1:NewNumOfNodes)]; NumOfNodes = NewNumOfNodes; Nodes = [Nodes(1:NumOfNodes,:)]; end case Node_Samples(2,2) NewNumOfNodes = Node_Samples(1,2); if (NewNumOfNodes > NumOfNodes) diff_size = NewNumOfNodes NumOfNodes; temp_Nodes = zeros (diff_size,7); temp_Nodes(:,N_Est_Cmn) = n_init; temp_Nodes(:,CW_Cmn) = 5 n_init + 15; %%% Initialize CW sizes %%% temp_Nodes(:,DC_Cmn) = DC; %%%%% Initialize DC values %%%%% for Node Num = 1:diff_size %% Initialize BackOff Conut for each nodes %% temp_Nodes(NodeNum,BC_Cmn) = fix(temp_Nodes(NodeNum, CW_Cmn)*rand(1,1)); end 93

PAGE 94

Each_Count = [Each_Count zeros(Each_N_H istory+3,diff_size)]; Nodes = [Nodes;temp_Nodes]; else Each_C ount = [Each_Count(:, 1:NewNumOfNodes)]; NumOfNodes = NewNumOfNodes; Nodes = [Nodes(1:NumOfNodes,:)]; end case Node_Samples(2,3) NewNumOfNodes = Node_Samples(1,3); if (NewNumOfNodes > NumOfNodes) diff_size = NewNumOfNodes NumOfNodes; temp_Nodes = zeros (diff_size,7); temp_Nodes(:,N_Est_Cmn) = n_init; temp_Nodes(:,CW_Cmn) = 5 n_init + 15; %%%% Initialize CW sizes %%%% temp_Nodes(:,DC_Cmn) = DC; %%%%% Initialize DC values %%%%% for Node Num = 1:diff_size %% Initialize BackOff Conut for each nodes %% temp_Nodes(NodeNum,BC_Cmn) = fix(temp_Nodes(NodeNum, CW_Cmn)*rand(1,1)); end Each_Count = [Each_Count zeros(Each_N_H istory+3,diff_size)]; NumOfNodes = NewNumOfNodes; Nodes = [Nodes;temp_Nodes]; else Each_C ount = [Each_Count(:, 1:NewNumOfNodes)]; NumOfNodes = NewNumOfNodes; Nodes = [Nodes(1:NumOfNodes,:)]; end case Node_Samples(2,4) NewNumOfNodes = Node_Samples(1,4); if (NewNumOfNodes > NumOfNodes) diff_size = NewNumOfNodes NumOfNodes; temp_Nodes = zeros (diff_size,7); temp_Nodes(:,N_Est_Cmn) = n_init; temp_Nodes(:,CW_Cmn) = 5 n_init + 15; %%%% Initialize CW sizes %%%% temp_Nodes(:,DC_Cmn) = DC; %%%% Initialize DC values %%%% for Node Num = 1:diff_size %% Initialize BackOff Conut for each nodes %% temp_Nodes(NodeNum,BC_Cmn) = fix(temp_Nodes(NodeNum, CW_Cmn)*rand(1,1)); end Each_Count = [Each_Count zeros(Each_N_H istory+3,diff_size)]; NumOfNodes = NewNumOfNodes; Nodes = [Nodes;temp_Nodes]; else Each_C ount = [Each_Count(:, 1:NewNumOfNodes)]; NumOfNodes = NewNumOfNodes; 94

PAGE 95

Nodes = [Nodes(1:NumOfNodes,:)]; end case Node_Samples(2,5) NewNumOfNodes = Node_Samples(1,5); if (NewNumOfNodes > NumOfNodes) diff_size = NewNumOfNodes NumOfNodes; temp_Nodes = zeros (diff_size,7); temp_Nodes(:,N_Est_Cmn) = n_init; temp_Nodes(:,CW_Cmn) = 5 n_init + 15; %%% Initialize CW sizes %%% temp_Nodes(:,DC_Cmn) = DC; %%%% Initialize DC values %%%% for Node Num = 1:diff_size %% Initialize BackOff Conut for each nodes %% temp_Nodes(NodeNum,BC_Cmn) = fix(temp_Nodes(NodeNum, CW_Cmn)*rand(1,1)); end Each_Count = [Each_Count zeros(Each_N_H istory+3,diff_size)]; NumOfNodes = NewNumOfNodes; Nodes = [Nodes;temp_Nodes]; else Each_C ount = [Each_Count(:, 1:NewNumOfNodes)]; NumOfNodes = NewNumOfNodes; Nodes = [Nodes(1:NumOfNodes,:)]; end end Each_Count(1,:) = Each_Count(1,:) + 1; ChkNodes = find(Nodes(:,BC_Cmn) == 0); %%%%% Check whose BC = 0 %%%%%%% [ChkCol temp] = size(ChkNodes); %%%%%%% ('COLLISION' or 'SUCCESS) %%%%%%%%%%%%%%% if (ChkCol >= 1) if (ChkCol > 1) T_Col_Cnt = T_Col_Cnt + 1; for coll_check = 1:ChkCol Nodes(ChkNodes(coll_check,1),Coll_Cmn) = Nodes(ChkNodes(coll_check,1),Coll_Cmn) + 1; end else T_Suc_Cnt = T_Suc_Cnt + 1; Nodes(ChkNodes,SU C_Cnt) = Nodes(ChkNodes,SUC_Cnt) + 1; end for Node_element = 1:NumOfNodes, if (Nodes(Node_eleme nt,BC_Cmn) == 0) %%%%% For Colliding Nodes %%%%% if (Each_Count(1 ,Node_element) >= Nodes(Node_element,CW_Cmn)) Each_Count (2,Node_element) = Each_Count(2,Node_element) + 1; 95

PAGE 96

Each_Count(3,Node_element) = fix(((Nodes(Node_element,Busy_Cmn)+Nodes(Node_element,Coll_Cmn))*(2*Nodes(Node_ele ment,CW_Cmn)-DC))/(2*Each_Count(1,Node_element))+1); if (Each_Count(2,Node_element ) <= Each_N_History) Nodes(N ode_element,N_Est_Cmn) = fix(Each_Count(3,Node_element)); else temp_cumsum = cumsum(Each_Count(4:13,Node_element)); Nodes(Node _element,N_Est_Cmn) = fix(al*E ach_Count(3,Node_element) + (1al)*(temp_cumsum(Each_N_History,1)/Each_N_History)); end Nodes(Node_ele ment,CW_Cmn) = fix(5*Nodes(N ode_element,N_Est_Cmn)+15); Nodes(Node_eleme nt,BC_Cmn) = fix(Nodes(Node_element,CW_Cmn)*rand(1)); Each_Count(1,Node_element) = 0; Nodes(Node_element,Busy_Cmn) = 0; Nodes(Node_element,DC_Cmn) = DC; else Nodes(Node_e lement,DC_Cmn) = DC; %%%Res et Their DC Counter %%%% Nodes(Node_eleme nt,BC_Cmn) = fix(Nodes(Node_element,CW_Cmn)*rand(1)); end else %% For Other Nodes (Not Colliding Nodes at that Contention Period) %% Nodes(Node_eleme nt,DC_Cmn) = Nodes(Node_element,DC_Cmn) 1; Nodes(Node_eleme nt,BC_Cmn) = Nodes(Node_element,BC_Cmn) 1; Nodes(Node_eleme nt,Busy_Cmn) = Nodes(Node_element,Busy_Cmn) + 1; if (Nodes(Node_element,DC_Cmn) < 0) Nodes(Node_eleme nt,BC_Cmn) = fix(Nodes(Node_element,CW_Cmn)*rand(1)); end end % End IF % end % End FOR % else %%%%%%%% %%%%%('IDLE') %%%%%%%%%%%%%% T_Idle_Cnt = T_Idle_Cnt + 1; %T_Idle_Cnt(Node_inde x) = T_Idle_Cnt(Node_index) + 1; Nodes(:,BC_Cmn) = Nodes(:,BC_Cmn) 1; %Nodes(:,BC0_Cmn) = Nodes(:,BC0_Cmn) + 1; end select_one = fix((N umOfNodes-1)*rand(1)+1); temp_N_Est(NumOfSim) = Nodes(select_one,N_Est_Cmn); if (NumOfSim == 20000) plot(1:NumOfNode s,2./(2.*Nodes(:,CW_Cmn)-DC)); end end MAC_Eff = (T_Suc_Cnt*Tsuc_vec)/(T_Suc_Cnt*(Tsuc_vec+TACK+TSIFS+TDIFS) +... T_Col_Cnt*(Tcol_vec+TDIFS) + T_Idle_Cnt*Tslot); 96

PAGE 97

plot(1:T_NumOfSim,temp_N_Est); grid on; hold on plot(Node_Samples(2,1):Node_Samples (2,2)-1, Node_Samples(1,1),'r-',... Node_Samples(2,2):Node_Sam ples(2,3)-1, Node_Samples(1,2),'r-',... Node_Samples(2,3):Node_Sam ples(2,4)-1, Node_Samples(1,3),'r-',... Node_Samples(2,4):Node_Sam ples(2,5)-1, Node_Samples(1,4),'r-',... Node_Samples(2,5):T_NumO fSim, Node_Samples(1,5),'r-'); hold off Proposed Optimal CW based Backoff Algorit hm with Estimation Scheme under Hidden Node Conditions clear all; close all; Node_Samples = [5:5:50 60:10:100]; CW_index = [32 64 128 256 512 1025]; CWs = 3; % Contention Windows for i = 1:6 CWs = CW_index(i); for Node_index = 1:max(size(Node_Samples)), % How many nodes in the simulation NumOfNodes = Node_Samples(Node_index); BPC = 1; % BackOff Procedure Counter DC = 3; % Deferral Counter T_NumOfSim = 20000; % Total Number of Simulation BusyCount = 0; AvgNumOfDCrst = 0; Inside_Count = 0; n_init = 5; CW_Cmn = 1; DC_Cmn = 2; BC_Cmn = 3; Busy_Cmn = 4; SUC_Cnt = 5; N_Est_Cmn = 6; Coll_Cmn = 7; Count_Cmn = 8; GroupID = 5; PercentageOfHiddens = 20; T_Col_Cnt = zeros(1,max(size(Node_Samples))); % Initialize Total Number of Collisions 97

PAGE 98

T_Suc_Cnt = zeros(1,max(size(Node_Sampl es))); % Initialize To tal Number of Success T_Idle_Cnt = zeros(1,max(size(Node_Sam ples))); % Initialize Total Number of Idles T_Suc_Cnt_Hidden = zeros(1,max(size(Node_Samples))); al = 0.8; Each_N_History = 10; Each_Count = zeros(E ach_N_History+3,NumOfNodes); Each_Count(3,:) = n_init; Frame_Count = 39; Tslot = 20; % Slot Time TSIFS = 0;%.5*Tslot; % SIFS Time TDIFS = 0;%2.5*Tslot; % DIFS Time TACK = 0;%5*Tslot; % ACK Time Tsuc_vec = 40*Tslot; % Successful Packet Transmission Time Tcol_vec = Tsuc_vec; % Colliding Packet Transmission Time Count = 0; Nodes = zeros(NumOfNodes,8); Nodes(:,N_Est_Cmn) = n_init; Nodes(:,CW_Cmn) = 5 n_init + 10; %%%%%% Initialize CW sizes %%%%%% Nodes(:,DC_Cmn) = DC; %%%%%% Initia lize DC values %%%%%% Nodes(:,SUC_Cnt) = 0; %% %% Initialize Success Count for check fairness %%%% Nodes(:,Busy_Cmn) = 0; %%%%%% Initialize Total # of DC Resets %%%%%% for NodeNum = 1:NumOfNodes %%% Initialize BackOff Conut for each nodes %%% Nodes(NodeNum,BC_Cmn) = floor(Nodes(NodeNum,CW_Cmn)*rand(1,1)); end for i = 1:NumOfNodes-1, TotalPath = TotalPath + i; end NumOfHiddenN = round(TotalP ath (PercentageOfHiddens/100)); TempHnodes = [1:NumOfNodes]; state = randint(1,1,[1,100]); neworder = randintrlv(TempHnodes,state); for i = 1:floor(NumOfNodes/5), Hnodes(i,1) = neworder(1,(2*i-1)); Hnodes(i,2) = neworder(1,(2*i)); end HnodeVecs = Hnodes(:)'; CWs = 5*NumOfNodes + 10; %%%%% INITIAL STATE %%%%%% 98

PAGE 99

Nodes(:,BPC_Cmn) = BPC; Nodes(:,DC_Cmn) = DC; Nodes(:,BC_Cmn) = round((CWs-1)*rand(NumOfNodes,1)); Nodes(:,Count_Cmn) = Frame_Count; Nodes(:,Index_Cmn) = [1:NumOfNodes]'; %%%% Setting some hidden no des, 20% of total nodes, %%%% %%%% Approximately 10 % of total path. %%%%%%%%%%% HnodeVecs = round(1 + (NumOfNodes 1) rand(1, NumOfHiddenN)); %%%% HnodeVecs : randomly selected. %%%% %%%% Hnodes : total path s between the hidden nodes %%%% for i = 2:max(size(HnodeVecs)), RNDTest = 1; while (RNDTest ~= 0) atemp1=HnodeVecs(1,1:i-1); atemp2=HnodeVecs(1,i+1:max(size(HnodeVecs))); atempfinal=[atemp1,atemp2]; RNDagain = find(HnodeVecs(1,i) == atempfinal(1,:)); [temp, RNDTest] = size(RNDagain); if (RNDTest ~= 0) HnodeVecs (1,i) = round(1+ (NumOfNodes 1) rand(1)); else RNDTest = 0; end end end tempcount = 0; for i = 1:max(size(HnodeVecs)) 1, for j = i + 1:max(size(HnodeVecs)), ttemp = [HnodeVecs(1,i),HnodeVecs(1,j)]; tempcount = tempcount + 1; Hnodes(tempcount,:) = ttemp; end end %%%%%%%%%%%%%% DESTINAT ION Node setting %%%%%%%%%%%%% Nodes(:,Destin_Cmn) = r ound(1+(NumOfNodes-1)*r and(NumOfNodes,1)); for NodeNum = 1:NumOfNodes %Nodes(NodeNum,Destin_Cmn) = NodeNum; while (NodeNum == Nodes(NodeNum,Destin_Cmn)) Nodes(NodeNum,De stin_Cmn) = round(1 + (Num OfNodes-1)*rand(1)); 99

PAGE 100

end end for NodeNum = 1:NumOfHiddenN while (Nodes(Hnodes(NodeNum,1),Destin_Cmn) == Hnodes(NodeNum,2) || Nodes(Hnodes(NodeNum,1),Destin _Cmn) == Hnodes(NodeNum,1)) Nodes(Hnodes(NodeNu m,1),Destin_Cmn) = round(1+(NumOfNodes-1)*rand(1)); end while (Nodes(Hnodes(N odeNum,2),Destin_Cmn) == Hnodes(NodeNum,1) || Nodes(Hnodes(NodeNum,2),Destin _Cmn) == Hnodes(NodeNum,2)) Nodes(Hnodes(NodeNu m,2),Destin_Cmn) = round(1+(NumOfNodes-1)*rand(1)); end end tempcount = 0; for i = 1:NumOfNodes-1, for j = i+1: NumOfNodes tempcount = tempcount + 1; BeforeHnodes(tempcount,:) = [i,j]; end end tempTotalPath = [1:TotalPath]; state = randint(1,1,[1,100]); neworder = randintrlv(tempTotalPath,state); for i = 1:NumOfHiddenN, Hnodes(i,:) = BeforeHnodes(neworder(i),:); end tempcount = 0; HHnodes = Hnodes(:)'; for i = 1:max(size(HHnodes)), TEMPHHnodes = HHnodes(i+1:max(size(HHnodes))); [temp_same,same_eleme nt] = size(find(HHnodes( i) == TEMPHHnodes)); if(same_element == 0) tempcount = tempcount + 1; HnodeVecs(1,tempcount) = HHnodes(i); end end for NumOfSim = 1:T_NumOfSim, Each_Count(1,:) = Each_Count(1,:) + 1; ChkNodes = find(Nodes(:,BC_Cmn) == 0); [ChkCol temp] = size(ChkNodes); 100

PAGE 101

if (ChkCol >= 2) %%%%%%%%%%%% ('COLLISION') %%%%%%%%%%%% T_Col_Cnt(Node _index) = T_Col_Cnt(Node_index) + 1; Nodes(ChkNodes, Coll_Cmn) = Nodes(ChkNodes, Coll_Cmn) + 1; for Node_element = 1:NumOfNodes, if (Nodes(Node_element,BC_Cmn) == 0) if (Each_C ount(1,Node_element) > Node s(Node_element,CW_Cmn)) Eac h_Count(2,Node_element) = Eac h_Count(2,Node_element) + 1; Each_C ount(4:13,Node_element) = Eac h_Count(3:12,Node_element); Each_Count(3,Node_element) = fix(((Nodes(Node_element,Busy_Cmn)+Nodes(Node_element,Coll_Cmn))*(2*Nodes(Node_ele ment,CW_Cmn)-DC))/(2*Each_Count(1,Node_element))+1); if (Each_Count(2,Node_element) <= Each_N_History) Node s(Node_element,N_Est_Cmn) = fi x(Each_Count(3,Node_element)); else temp_cumsum = cumsum(Each_Count(4:13,Node_element)); Nodes( Node_element,N_Est_Cmn) = fix( al*Each_Count(3,Node_element) + (1-al)*(temp_cumsum(Each_N_H istory,1)/Each_N_History)); end Nodes(Node_element,CW_Cmn) = fix(5*Nodes(Node_element,N_Est_Cmn)+10); Nodes(Node_element,BC_Cmn) = fix(Nodes(Node_element,CW_Cmn)*rand(1)); Each_Count(1,Node_element) = 0; Nodes(Node_element,Coll_Cmn) = 0; Nodes(Node_element,Busy_Cmn) = 0; Nodes(Node_element,DC_Cmn) = DC; else Nodes(Node_element,DC_Cmn) = DC; %%% Reset Their DC Counter %%%% Nodes(N ode_element,Busy_Cmn) = Nodes(Node_element,Busy_Cmn) + 1; Nodes(Node_e lement,CW_Cmn) = fix(5 Nodes(Node_element,N_Est_Cmn) + 35); Nodes(Node_element,BC_Cmn) = fix(Nodes(Node_element,CW_Cmn)*rand(1)); end else Nodes(Node_e lement,DC_Cmn) = Nodes(Node_element,DC_Cmn) 1; Nodes(Node_e lement,BC_Cmn) = Nodes(Node_element,BC_Cmn) 1; Nodes(Node_element,Busy_Cmn) = Nodes(Node_element,Busy_Cmn) + 1; if (Nodes(Node_element,DC_Cmn) < 0) Nodes(Node_element,DC_Cmn) = DC; Nodes(N ode_element,Busy_Cmn) = Nodes(Node_element,Busy_Cmn) + 1; Nodes(Node_element,BC_Cmn) = fix(Nodes(Node_element,CW_Cmn)*rand(1)); end 101

PAGE 102

end end elseif (ChkCol == 1) %%%%%%%%%%%('S UCCESS')%%%%%%%%%%% [ChkHORow Chk HOCol] = find(HnodeVecs(:,:) == ChkNodes); [ChkHiddenOpen temp] = size(ChkHORow); HiddenPair = find(Nodes(HnodeVecs(1,:),BC_Cmn) == 0);. [InsideHidden,temp] = size(find(Nodes(HnodeVecs(1,:),BC_Cmn) == 0)); if (ChkHOCol == 1) PairN = 2; else PairN = 1; end if (InsideHidden == 0) T_Suc_Cnt( Node_index) = T_Suc_Cn t(Node_index) + 1; Nodes(ChkNode s,SUC_Cnt) = Nodes(ChkNodes,SUC_Cnt) + 1; for Node_element = 1:NumOfNodes, if (Nodes(Node_element,BC_Cmn) == 0) if (Each_Count(1,Node_element) > Nodes(Node_element,CW_Cmn)) Each_Count(2,Node_element) = Each_Count(2,Node_element) + 1; E ach_Count(4:13,Node_element) = E ach_Count(3:12,Node_element); Each_Count(3,Node_element) = fix(((Nodes(Node_element,Busy_Cmn)+Nodes(Node_element,Coll_Cmn))*(2*Nodes(Node_ele ment,CW_Cmn)-DC))/(2*Each_Count(1,Node_element))+1); if (Each_Count(2,Node_ele ment) <= Each_N_History) Nodes(Node_element,N_Est_Cmn) = fix(Each_Count(3,Node_element)); else temp_cumsum = cumsum(Each_Count(4:13,Node_element)); Nodes(Node_element,N_Est_Cmn) = fix(al*Each_Count(3,Node_element) + (1-al)*(temp_cumsum(Each_N_H istory,1)/Each_N_History)); end Nodes(Node_element,CW_Cmn) = fix(5*Nodes(Node_element,N_Est_Cmn)+10); Nodes(Node_element,BC_Cmn) = fix(Nodes(Node_element,CW_Cmn)*rand(1)); Each_Count(1,Node_element) = 0; Nodes(Node_element,Coll_Cmn) = 0; Nodes(Node_element,Busy_Cmn) = 0; Nodes(Node_element,DC_Cmn) = DC; else Nodes( Node_element,DC_Cmn) = DC; %% % Reset Their DC Counter %%% Nodes(Node_element,Busy_Cmn) = Nodes(Node_element,Busy_Cmn) + 1; 102

PAGE 103

Nodes( Node_element,CW_Cmn) = fix(5 Nodes(Node_element,N_Est_Cmn) + 35); Nodes(Node_element,BC_Cmn) = fix(Nodes(Node_element,CW_Cmn)*rand(1)); end else %% For Other Nodes (Not Colliding Nodes at that Contention Period) %% Nodes(N ode_element,DC_Cmn) = Nodes(Node_element,DC_Cmn) 1; Nodes(Node_element,BC_Cmn) = Nodes(Node_element,BC_Cmn) 1; Nodes(N ode_element,Busy_Cmn) = Nodes(Node_element,Busy_Cmn) + 1; if (Nodes(Node_element,DC_Cmn) < 0) Nodes(Node_element,DC_Cmn) = DC; Nodes(Node_element,Busy_Cmn) = Nodes(Node_element,Busy_Cmn) + 1; Nodes(Node_element,BC_Cmn) = fix(Nodes(Node_element,CW_Cmn)*rand(1)); end end % End IF % end else [BCzeroRow,BCzer oCol] = find(HnodeVecs(Hid denPair) == Hnodes(:,:)); temp_ BCzeroRowCol = [BCzeroRow,BCzeroCol]; if BCzeroCol == 1 PairBCzeroCol = 2; else PairBCzeroCol = 1; end while (Nodes( Hnodes(BCzeroRow,BCzeroCol),BC_Cmn) == 0 || Nodes(Hnodes(BCzeroRow,Pair BCzeroCol),BC_Cmn) == 0) if (N odes(Hnodes(BCzeroRow,BCzeroCol),BC_Cmn) == 0 && Nodes(Hnodes(BCzeroRow,Pair BCzeroCol),BC_Cmn) ~= 0) Nodes(Hnodes(BCzeroRow,Pai rBCzeroCol),BC_Cmn) = Nodes(Hnodes(BCzeroRow,Pai rBCzeroCol),BC_Cmn) -1; Nodes(Hnodes(BCzeroRow,B CzeroCol),Count_Cmn) = Nodes(Hnodes(BCzeroRow,BC zeroCol),Count_Cmn) 1; if (Nodes(Hnodes(BCzeroRow,BC zeroCol),Count_Cmn) == 0) Nodes(Hnodes(BCzeroRow,BCzeroCol),SUC_Cnt) = Nodes(Hnodes(BCzeroRow,BC zeroCol),SUC_Cnt) + 1; Nodes(Hnodes(BCzeroRow,BCzeroCol),Coll_Cmn) = Nodes(Hnodes(BCzeroRow,BC zeroCol),Coll_Cmn) + 1; Node s(Hnodes(BCzeroRow,BCzeroCol),BC _Cmn) = round((CWs-1)*rand(1)); Nodes(Hnodes(BCzeroRow, BCzeroCol),Count_Cmn) = 39; end 103

PAGE 104

elseif (Nodes(Hnodes(BCzeroRow,BC zeroCol),BC_Cmn) == 0 && Nodes(Hnodes(BCzeroRow,Pair BCzeroCol),BC_Cmn) == 0) Nodes(Hnodes(BCzeroRow,B CzeroCol),Count_Cmn) = Nodes(Hnodes(BCzeroRow,BC zeroCol),Count_Cmn) 1; Nodes(Hnodes(BCzeroRow,Pai rBCzeroCol),Count_Cmn) = Nodes(Hnodes(BCzeroRow,PairBC zeroCol),Count_Cmn) 1; if (Nodes(Hnodes(BCzeroRow,BC zeroCol),Count_Cmn) == 0) Nodes(Hnodes(BCzeroRow,BCzeroCol),SUC_Cnt) = Nodes(Hnodes(BCzeroRow,BC zeroCol),SUC_Cnt) + 1; Nodes(Hnodes(BCzeroRow,BCzeroCol),Coll_Cmn) = Nodes(Hnodes(BCzeroRow,BC zeroCol),Coll_Cmn) + 1; Node s(Hnodes(BCzeroRow,BCzeroCol),BC _Cmn) = round((CWs-1)*rand(1)); Nodes(Hnodes(BCzeroRow, BCzeroCol),Count_Cmn) = 39; end if (Nodes(Hnodes(BCzeroRow,Pair BCzeroCol),Count_Cmn) == 0) Nodes(Hnodes(BCzeroRow,P airBCzeroCol),SUC_Cnt) = Nodes(Hnodes(BCzeroRow,PairBCzeroCol),SUC_Cnt) + 1; Nodes(Hnodes(BCzeroRow,P airBCzeroCol),Coll_Cmn) = Nodes(Hnodes(BCzeroRow,PairBCzeroCol),Coll_Cmn) + 1; Nodes(Hnodes(BCzeroRow,PairBC zeroCol),BC_Cmn) = round((CWs1)*rand(1)); Nodes(Hnodes(BCzeroRow,PairBCzeroCol),Count_Cmn) = 39; end elseif (Nodes(Hnodes(BCzeroRow,BC zeroCol),BC_Cmn) ~= 0 && Nodes(Hnodes(BCzeroRow,PairBCzeroCol),BC_Cmn) == 0) Nodes(Hnodes(BCzeroRow,Pai rBCzeroCol),Count_Cmn) = Nodes(Hnodes(BCzeroRow,PairBCzeroCol),Count_Cmn) -1; Nodes(Hnodes(BCzeroRow ,BCzeroCol),BC_Cmn) = Nodes(Hnodes(BCzeroRow,BC zeroCol),BC_Cmn) 1; if (Nodes(Hnodes(BCzeroRow,Pair BCzeroCol),Count_Cmn) == 0) Nodes(Hnodes(BCzeroRow,P airBCzeroCol),SUC_Cnt) = Nodes(Hnodes(BCzeroRow,PairBCzeroCol),SUC_Cnt) + 1; Nodes(Hnodes(BCzeroRow,P airBCzeroCol),Coll_Cmn) = Nodes(Hnodes(BCzeroRow,PairBCzeroCol),Coll_Cmn) + 1; Nodes(Hnodes(BCzeroRow,PairBC zeroCol),BC_Cmn) = round((CWs1)*rand(1)); Nodes(Hnodes(BCzeroRow,PairBCzeroCol),Count_Cmn) = 39; end else Nodes(Hnodes(BCzeroRow,BC zeroCol),Count_Cmn) = 39; N odes(Hnodes(BCzeroRow,PairBC zeroCol),Count_Cmn) = 39; end Inside_Count = Inside_Count + 1; end % End of While Inside_Count = 0; 104

PAGE 105

%%%%%%%%%% From here, other open node operation %%%%%%%%%%%% for Node_element = 1:NumOfNodes if (Node_eleme nt ~= Hnodes(BCzeroRow,BCzeroCol) | Node_element ~= Hnodes(BCzeroRow,P airBCzeroCol)) Nodes(N ode_element,DC_Cmn) = Nodes(Node_element,DC_Cmn) 1; Nodes(Node_element,BC_Cmn) = Nodes(Node_element,BC_Cmn) 1; Nodes(N ode_element,Busy_Cmn) = Nodes( Node_element,Busy_Cmn) + 1; if (Nodes(Node_element,DC_Cmn) < 0) Nodes(Node_element,DC_Cmn) = DC; Nodes(Node_element,Busy_Cmn) = Nodes(Node_element,Busy_Cmn) + 1; Nodes(Node_element,BC_Cmn) = fix(Nodes(Node_element,CW_Cmn)*rand(1)); end end end T_Suc_Cnt_Hidden(N ode_index) = T_Suc_Cnt_Hi dden(Node_index) + mok; mok = 0; end else %%%%%%%%%%%('IDLE' )%%%%%%%%%%%%%% T_Idle_Cnt(Node_ index) = T_Idle_Cnt (Node_index) + 1; Nodes(:,BC_Cmn) = Nodes(:,BC_Cmn) 1; end end MAC_Eff(Node_index) = (T_S uc_Cnt(Node_index)*Tsuc_vec)/... (T_Suc_Cnt(Node_inde x)*(Tsuc_vec+TACK+TSIFS+TDIFS) + ... ((T_Col_Cnt(Node_index) + T_Su c_Cnt_Hidden(Node_index))*(Tcol_vec+TDIFS)) + T_Idle_Cnt(Node_index)*Tslot); (T_Col_Cnt(Node_index)+T_Suc_Cnt_ Hidden(Node_index))*(T col_vec+TDIFS) + T_Idle_Cnt(Node_index)*Tslot); MAC_Eff(Node_index) = 40*T_Suc _Cnt(Node_index)/Tslot*T_NumOfSim; end plot(Node_Samples, MAC_Eff,'ks-');grid on; hold on; %axis([1 100 0.3 1]) hold on; end 105

PAGE 106

LIST OF REFERENCES [1] HomePlug 1.0 Specification, HomePlug Powerline Alliance, Jun. 2001. [2] S. Baig and N. D. Gohar, "A discrete multitone transceiver at the heart of the PHY layer of an in-home power line co mmunication local area network," IEEE Commun. Mag., vol. 41, pp. 48-53, 2003. [3] Y. J. Lin, H. A. Latchman, R. E. Newm an, and S. Katar, "A comparative performance study of wireless and power line networks," IEEE Commun. Mag., vol. 41, pp. 54-63, 2003. [4] A. Leon-Garcia and I. Widjaja, Communication Networks: McGraw-Hill, 2004. [5] Y. Xiao and J. Rosdahl, "Throughput and delay limits of IEEE 802.11," IEEE Commun. Lett. vol. 6, pp. 355-357, 2002. [6] J. Barnes, "A physical multi-path model fo r power distribution network propagation," in Proc. Int. Symp. Powerline Communications and its Applications Tokyo, Japan, 1998.pp. 76-89. [7] J. A. C. Bingham, "Multicarrier Modulation for Data-Transmission an Idea Whose Time Has Come," IEEE Commun. Mag., vol. 28, pp. 5-14, 1990. [8] G. Bianchi, "Performance analysis,of the IEEE 802.11 distributed coordination function," IEEE J. Select. Areas Commun. vol. 18, pp. 535-547, 2000. [9] IEEE standard for Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, ISO/IEC 802-11: 1999(E), Aug. 1999. [10] L. Kleinrock and F. A. Tobagi, "Packet Switching in Ra dio Channels .1. Carrier Sense Multiple-Access Modes and Their Throughput-Delay Characteristics," IEEE Trans. Commun. vol. 23, pp. 1400-1416, 1975. [11] F. A. Tobagi and L. Kleinrock, "Packet Switching in Radio Channels .4. Stability Considerations and Dynamic Control in Carrier Sense Multiple Access," IEEE Trans. Commun. vol. 25, pp. 1103-1119, 1977. [12] F. Cali, M. Conti, and E. Gregori, "Dynamic tuning of the IEEE 802.11 protocol to achieve a theoreti cal throughput limit," IEEE-ACM Trans. Network. vol. 8, pp. 785-799, 2000. [13] F. Cali, M. Conti, and E. Gregor i, "IEEE 802.11 protocol: Design and performance evaluation of an adaptive backoff mechanism," IEEE J. Select. Areas Commun. vol. 18, pp. 1774-1786, 2000. [14] G. Bianchi, "IEEE 802.11-sa turation throughput analysis," IEEE Commun. Lett. vol. 2, pp. 318-320, 1998. [15] G. Sharma, A. Ganesh, and P. Key, "Performance Analysis of Contention Based Medium Access Control Protocols," in Proc. INFOCOM 2006. 25th IE EE Int. Conf. Computer Commun. 2006.pp. 1-12. [16] H. Wu, Y. Peng, K. Long, S. Cheng, a nd J. Ma, "Performance of reliable transport protocol over IEEE 802.11 wi reless LAN: analysis and enhancement," in Proc. INFOCOM 2002. Twenty-First Annu. Joint Conf. IEEE Comput. Commun. Soc. 2002.pp. 599-607 vol.2. [17] P. Chatzimisios, A. C. Boucouvalas, a nd V. Vitsas, "IEEE 802.11 packet delay-a finite retry limit analysis," in Proc. IEEE GLOBECOM '03., 2003.pp. 950-954 Vol.2. [18] M. M. Carvalho and J. J. Garcia-Luna-A ceves, "Delay analysis of IEEE 802.11 in singlehop networks," in Proc. 11th IEEE Int. Conf. Netw. Protocols 2003.pp. 146-155. 106

PAGE 107

[19] G. Wang, Y. Shu, L. Zhang, and O. W. W. Yang, "Delay analysis of the IEEE 802.11 DCF," in Proc. 14th IEEE PIMRC 2003, 2003.pp. 1737-1741 vol.2. [20] P. Chatzimisios, A. C. Boucouvalas, and V. Vitsas, "IEEE 802.11 Wireless LANs: Performance Analysis and Protocol Refinement," EURASIP J. Wireless Commun. Netw., vol. 2005, pp. 67-78, 2005. [21] C. H. Foh and J. W. Tantra, "Comme nts on IEEE 802.11 saturation throughput analysis with freezing of backoff counters," IEEE Commun. Lett. vol. 9, pp. 130-132, 2005. [22] G. Bianchi and I. Tinnirello, "Remarks on IEEE 802.11 DCF performance analysis," IEEE Commun. Lett. vol. 9, pp. 765-767, 2005. [23] Q. Ni, I. Aad, C. Barakat, and T. Turl etti, "Modeling and analysis of slow CW decrease IEEE 802.11 WLAN," in Proc. 14th IEEE PIMRC 2003 2003.pp. 1717-1721 vol.2. [24] V. Vitsas and A. C. Boucouvalas, "Perfo rmance analysis of the Advanced Infrared (AIr) CSMA/CA MAC protocol for wireless LANs," Wireless Networks vol. 9, pp. 495-507, 2003. [25] Y. Xiao, "A simple and effective priority scheme for IEEE 802.11," IEEE Commun. Lett. vol. 7, pp. 70-72, 2003. [26] R. Bruno, M. Conti, and E. Gregori, "Optimal capacity of p-persistent CSMA protocols," IEEE Commun. Lett. vol. 7, pp. 139-141, 2003. [27] Y. C. Tay and K. C. Chua, "A capacity analysis for the IEEE 802.11 MAC protocol," Wireless Networks vol. 7, pp. 159-171, 2001. [28] A. Kumar, E. Altman, D. Miorandi, and M. Goyal, "New insights from a fixed point analysis of single cell IEEE 802.11 WLANs," in Proc. INFOCOM 2005. 24th Ann. Joint Conf. IEEE Comput. Commun. Soc. 2005.pp. 1550-1561 vol. 3. [29] K. Medepalli and F. A. Tobagi, "Thr oughput analysis of IEEE 802.11 wireless LANs using an average cycle time approach," in Proc. IEEE Glob. Telecommun. Conf. 2005, 2005.pp. 5 pp. [30] X. Wang, "Performance modeling of IEEE 802.11 DCF using equilibrium point analysis," in Proc. 20th Int. Conf. AINA 2006. 2006.pp. 6 pp. [31] Y. Kwon, Y. Fang, and H. Latchman, "D esign of MAC protocols with fast collision resolution for wireless local area networks," IEEE Tran. Wireless Commun. vol. 3, pp. 793-807, 2004. [32] N.-O. Song, B.-J. Kwak, J. Song, and M. E. Miller, "Enhancement of IEEE 802.11 distributed coordination func tion with exponential increase exponential decr ease backoff algorithm," in Proc. IEEE Veh. Technol. Conf., Spr. 2003 2003.pp. 2775-2778 vol.4. [33] B. J. Kwak, N. O. Song, and L. E. Mille r, "Performance analysis of exponential backoff," IEEE-Acm Trans. on Networking vol. 13, pp. 343-355, 2005. [34] V. Bharghavan, A. Demers, S. Shenker, and L. Zhang, MACAW: a media access protocol for wireless LAN's : ACM SIGCOMM, Oct. 1994. [35] I. Aad, Q. Ni, C. Barakat, and T. Tu rletti, "Enhancing IEEE 8 02.11 MAC in congested environments," Comput. Commun., vol. 28, pp. 1605-1617, 2005. [36] H. Wu, S. Cheng, Y. Peng, K. Long, and J. Ma, "IEEE 802.11 distributed coordination function (DCF): analysis and enhancement," in Proc. IEEE Int. Conf. Commun., 2002 2002.pp. 605-609. [37] Q. X. Pang, S. C. Liew, J. Y. B. Lee, a nd V. C. M. Leung, "Performance evaluation of an adaptive backoff scheme for WLAN," Wireless Commun. Mob. Com., vol. 4, pp. 867879, 2004. 107

PAGE 108

[38] C. G. Wang, B. Li, and L. M. Li, "A ne w collision resolution mechanism to enhance the performance of IEEE 802.11 DCF," IEEE Tran. Veh. Tech., vol. 53, pp. 1235-1246, 2004. [39] P. A. Brown, "Power line communica tions past, present, and future," in Proc. 1999 ISPLC Sep. 1999.pp. 1-8. [40] Y. Kaizawa and G. Marubayashi, "N eeds for the power line communications," in Proc. 1998 ISPLC 1998.pp. 153-157. [41] E. Ziouva and T. Antonakopoulos, "C SMA/CA performance under high traffic conditions: throughput a nd delay analysis," Comput. Commun. vol. 25, pp. 313-321, 2002. [42] G. Bianchi, L. Fratta, and M. Oliveri, "Performance evaluation and enhancement of the CSMA/CAMAC protocol for 802.11 wireless LANs," in Proc. PIMRC 1996 Taipei, Taiwan, Oct. 1996.pp. 392-396. [43] M. H. Jung, M. Y. Chung, and T. J. Lee, "MAC throughput anal ysis of HomePlug 1.0," IEEE Commun. Lett. vol. 9, pp. 184-186, 2005. [44] K. Tripathi, J. Lee, H. Latchman, J. McNair, and S. Katar, "Contention Window based Parameter Selection to Improve Powerline MA C Efficiency for Large Number of Users," in Proc. IEEE ISPLC Orlando, FL, Mar. 2006. [45] G. Bianchi and I. Tinnirello, "Kalman Filter Estimation of the Number of Competing Terminals in an IEEE 802.11 network," in Proc. IEEE Infocom San FranCisco, CA, Apr. 2003. [46] J. Lee, K. Tripathy, H. Latchman, and J. McNair, "Population Based Adaptive Tuning of Constant Contention Window HomePlug1.0," Int. J. Power and Energy Sys. vol. 28, pp. 222-228, 2008. [47] J. Lee, K. Tripathy, and H. A. Latc hman, "Efficient High Speed Communications over Electrical Power lines for a Large Number of Users," in Proc. Int. Conf. Power and Energy Sys., Clearwater, Florida, Jan. 2007.pp. 195-200. [48] HomePlug AV Specification HomePlug Powerline Alliance, Dec. 2005. [49] S. Khurana, A. Kahol, and A. P. Jayasummana, "Effect of Hidden Terminals on the Performance of IEEE 802.11 MAC Protocol," in Proc. LCN '98 Oct. 1998.pp. 12-20. [50] Y. Kim, J. Yu, S. Choi, and K. Jang, "A novel hidden station de tection mechanism in IEEE 802.11 WLAN," IEEE Commun. Lett. vol. 10, pp. 608-610, 2006. [51] H. Zhai and Y. Fang, "A Solution to Hi dden Terminal Problem over a Single Channel in Wireless AD HOC Networks," in Proc. MILCOM 2006 Oct. 2006.pp. 1-7. [52] K. Xu, M. Gerla, and S. Bae, "How e ffective is the IEEE 802.11 RTS/CTS handshake in ad hoc networks," in Proc. IEEE Glob. Telecommun. Conf. 2002.pp. 72-76. [53] Y. J. Lin, H. A. Latchman, J. C. L. Liu, and R. E. Newman, "Periodic Contention-Free Multiple Access For Broadband Multimedia Powerline Communication Networks," in Proc. ISPLC 2005 Apr. 2005.pp. 121-125. [54] D. Johnson, C. Perkins, and J. Arkko, "Mobility Support in IPv6," RFC 3775, IETF, June, 2003. [55] V. Tsaoussidis and I. Matta, "Open issues on TCP for mobile computing," Wireless Commun. & Mobile Compu. vol. 2, pp. 3-20, 2002. [56] C. Perkins and D. Johnson, "Route optimizati on in Mobile IP," Internet Draft, IETF, Sep. 2001. [57] R. Koodli, "Fast Handovers for Mobile IPv6," RFC 4068, IETF, July, 2005. 108

PAGE 109

[58] D. Lee, C. Oh, S. Lee, J. Park, and K. Kim, "Design and analysis of the mobile agent preventing out-of-sequence," in Proc. ICOIN Jan, 1999. [59] D. S. Eom, M. Sugano, M. Murata, and H. Miyahara, "Performance improvement by packet buffering in mobile IP based networks," IEICE Trans. Commun. vol. E83B, pp. 2501-2512, 2000. [60] H. Balakrishnan, S. Seshan, and R. H. Katz, "Improving Reliable Transport and Handoff Performance in Cellular Wireless Networks," ACM Wireless Networks vol. 1, pp. 279289, 1995. [61] The Network Simulator (NS2), http://www.isi.edu/nsnam [62] B. Park, J. Lee, and H. A. Latchman, "An Efficient TCP Mechanism to Reduce Out-ofsequence Packets in Mobile IPv6," in Proc. IEEE Int. Conf. Adv. Commun. Tech. Feb. 2006.pp. 1115-1118. 109

PAGE 110

BIOGRAPHICAL SKETCH Jongdae Lee received his Bachelor of Scie nce degree in information and communication engineering at Daejeon Universi ty, Korea in 1999 and his M.S. in electrical and computer engineering at Portland State Univ ersity in 2003. He is currently working for his Ph.D. degree in electrical and computer engineeri ng at University of Florida. His research interests are in the area of medium access control protocols in wi reless local area networks and power line communications. Specifically, he is working on designing a next genera tion MAC protocol and hidden node problems in power line networking. Also he is focu sing on fast handover scheme for mobile IPv6.