<%BANNER%>

Analysis and Performance Enhancements in Power Line, Wireless and Mobile Networks

Permanent Link: http://ufdc.ufl.edu/UFE0041581/00001

Material Information

Title: Analysis and Performance Enhancements in Power Line, Wireless and Mobile Networks
Physical Description: 1 online resource (106 p.)
Language: english
Creator: Lee, Sung
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: homenetwork, lan, mobilenetwork, plc
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: This dissertation focuses on the study of the performance of Powerline Communication (PLC), wireless and mobile networks in conveying multimedia traffic such as Internet Protocol Television (IPTV). We provide a comprehensive protocol performance comparison for wireless and PLC in-home networks. Our result shows that the throughput of Homeplug AV is at least 18Mbps higher than that of IEEE 802.11 (g and n) because of improved design and the quasi-stationary channel characteristics of PLC. We also show that Homeplug AV outperforms other competing PLC technologies by as much as 17.5% for TCP and 77.7% for UDP at 77% of connections and in addition achieves 100% full home coverage in our performance evaluation. We also developed and used in this work a network protocol visualization tool, ?Visugator? for performance evaluation of network protocols. This tool has been made freely available for public use. For distributing multimedia traffic in mobile networks a mobility management protocol is an essential element of mobile computing including mobile IPTV. We developed a Fast Movement Detection and DAD (Duplicate Address Detection) (FMDD) scheme for Mobile IPv6 protocol, as well as two improved network-based Proxy Mobile IPv6 protocols over wireless local area network that reduce handover latency and signaling overhead. In the Mobile IPv6 with FMDD scheme, the neighbor cache is used for DAD to reduce the handover latency. To reduce the handover latency and packet loss of Proxy Mobile IPv6, simultaneous L2 & L3 handover procedures and a tunnelling scheme are used. With the help of an Inter Access Point Protocol (IAPP) we decrease the signaling overhead of Proxy Mobile IPv6 significantly. A mathematical analysis is provided to show the benefits of our proposed schemes. Based on our analytical results, the FMDD scheme with Mobile IPv6 can reduce the DAD latency by lookup delay in RAM from the 1 second to just several microseconds. Our proposed Proxy mobile IPv6 protocols also can reduce up to 12.7% of the handover latency, 29.56% of packet loss and 31.6% of signaling overhead compared with the conventional Proxy Mobile IPv6. The dissertation also presents a MAC scheduling scheme of Bluetooth to improve MAC efficiency.
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 Sung Lee.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Latchman, Haniph A.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2012-04-30

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041581:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041581/00001

Material Information

Title: Analysis and Performance Enhancements in Power Line, Wireless and Mobile Networks
Physical Description: 1 online resource (106 p.)
Language: english
Creator: Lee, Sung
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: homenetwork, lan, mobilenetwork, plc
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: This dissertation focuses on the study of the performance of Powerline Communication (PLC), wireless and mobile networks in conveying multimedia traffic such as Internet Protocol Television (IPTV). We provide a comprehensive protocol performance comparison for wireless and PLC in-home networks. Our result shows that the throughput of Homeplug AV is at least 18Mbps higher than that of IEEE 802.11 (g and n) because of improved design and the quasi-stationary channel characteristics of PLC. We also show that Homeplug AV outperforms other competing PLC technologies by as much as 17.5% for TCP and 77.7% for UDP at 77% of connections and in addition achieves 100% full home coverage in our performance evaluation. We also developed and used in this work a network protocol visualization tool, ?Visugator? for performance evaluation of network protocols. This tool has been made freely available for public use. For distributing multimedia traffic in mobile networks a mobility management protocol is an essential element of mobile computing including mobile IPTV. We developed a Fast Movement Detection and DAD (Duplicate Address Detection) (FMDD) scheme for Mobile IPv6 protocol, as well as two improved network-based Proxy Mobile IPv6 protocols over wireless local area network that reduce handover latency and signaling overhead. In the Mobile IPv6 with FMDD scheme, the neighbor cache is used for DAD to reduce the handover latency. To reduce the handover latency and packet loss of Proxy Mobile IPv6, simultaneous L2 & L3 handover procedures and a tunnelling scheme are used. With the help of an Inter Access Point Protocol (IAPP) we decrease the signaling overhead of Proxy Mobile IPv6 significantly. A mathematical analysis is provided to show the benefits of our proposed schemes. Based on our analytical results, the FMDD scheme with Mobile IPv6 can reduce the DAD latency by lookup delay in RAM from the 1 second to just several microseconds. Our proposed Proxy mobile IPv6 protocols also can reduce up to 12.7% of the handover latency, 29.56% of packet loss and 31.6% of signaling overhead compared with the conventional Proxy Mobile IPv6. The dissertation also presents a MAC scheduling scheme of Bluetooth to improve MAC efficiency.
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 Sung Lee.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Latchman, Haniph A.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2012-04-30

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041581:00001


This item has the following downloads:


Full Text

PAGE 1

1 ANALYSIS AND PERFORMANCE ENHANCEMENTS IN POWER LINE, WIRELESS AND MOBILE NETWORKS By SUNGUK LEE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Sunguk Lee

PAGE 3

3 To my p arents and my wife Hyuna Kim

PAGE 4

4 ACKNOWLEDGMENTS First of all I would like to express my deep appreciation for my academic advisor, Dr. Haniph A. Latchman for his patient guidance, encouragement and plentiful advice until I can successfully finish my Doctoral research. I would also like to thank the members of my PhD. committee, Dr. Arroyo Dr. Janise McNair, and Dr. Fitz Coy. I am grateful for their willingness to serve on my committee and their helpful advice. I thank my parents who have kept encouraging and inspiring me to pursue my dream in many ways with unconditional love. I have to thank my lovely wife, Hyuna Kim She has patiently and enthusiastically supported my study. I also thank my colleagues at Laboratory for Information Systems and Telecommunications (LIST) in ECE department f or their many helpful and friendly discussions that a lways gave me a new realiz ation.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ...................................................................................................... 4 LIST OF TABLES ................................................................................................................ 7 LIST OF FIGURES .............................................................................................................. 8 ABSTRACT ........................................................................................................................ 10 CHAPTER 1 INTRODUCTION ........................................................................................................ 12 Home Network ............................................................................................................ 12 Mobility Management Protocol ................................................................................... 14 2 C OMPARATIVE PERFORMANCE ANALYSIS OF LOCAL AREA NETWORK TECHNOLOGIES FOR MULTIMEDIA STREAMING IN HOME ............................... 16 Introduction ................................................................................................................. 16 Overview of Wireless LAN and Powerline communication ....................................... 16 IEEE 802.11x Wireless LAN ................................................................................ 17 Power Line Communication ................................................................................. 18 Experiment Setup ....................................................................................................... 22 Test of UDP and TCP Protocol .................................................................................. 25 3 A VISUALIZATION TOOL FOR NETWORKING PROTOCOL ANALYSIS .............. 32 Introduction ................................................................................................................. 32 Requirements .............................................................................................................. 34 Design and Implementation ........................................................................................ 35 Design of the EVGator User Interface ................................................................. 36 Configuring the EVGATOR .................................................................................. 40 Trace File Format ................................................................................................. 40 Zoom In/Out ...................................................................................................... 42 Experience .................................................................................................................. 43 Conclusion .................................................................................................................. 44 4 ENHANCED LIMITED ROUND ROBIN MECHANISM USING PRIORITY POLICY OVER BLUETOOTH NETWORK ................................................................ 45 Introduction ................................................................................................................. 45 MAC scheduling scheme of Bluetooth ....................................................................... 46 ELRR -PP Algorithm .................................................................................................... 49

PAGE 6

6 Performance Evaluation ............................................................................................. 51 Conclusion .................................................................................................................. 54 5 EFFICIENT M O BIL ITY MANAGEMENT SCHEME OVER WIRELESS LOCAL AREA NETWORK ....................................................................................................... 56 Introduction ................................................................................................................. 56 Enhanced Handover Scheme of MIPv6 with Fast Movement Detection & DAD ..... 57 Overview of MobileIPv6 (MIPv6) ......................................................................... 57 Proposed MIPv6 Handover Scheme with FMDD ................................................ 60 Movement Detection using Stored RA message ......................................... 61 DAD configuration and Lookup Algorithm .................................................... 62 Efficient PMIPv6 Handover Scheme over Wireless Local Area Network ................. 64 Overview of Proxy Mobile IPv6 (PMIPv6) ........................................................... 64 Proposed PMIPv6 protocol 1 ............................................................................... 66 L2 Handover Procedure ................................................................................ 66 L3 Handover Procedure ................................................................................ 67 Proposed P MIPv6 Protocol 2 .............................................................................. 71 Analytical model .......................................................................................................... 73 Performance Analysis ................................................................................................. 77 Analysis of Handover Latency and Packet Loss ................................................. 78 Mobile IPv6. ................................................................................................... 79 Proxy Mobile IPv6 ......................................................................................... 81 Proposed PMIPv6 Protocol 1 ........................................................................ 82 Analysis of Signaling Cost ................................................................................... 83 Mobile IPv6 Protocol ..................................................................................... 85 Proxy Mobile IPv6 protocol ........................................................................... 86 Proposed PMIPv6 Protocol 2 ........................................................................ 87 Numeri cal Result and Comparison ............................................................................ 88 Numerical Result of MIPv6 with FMDD ........................................................ 89 Numerical Result of Proposed PMIPv6 Scheme 1 ...................................... 92 Numerical Result of Proposed PMIPv6 scheme 2 ....................................... 97 Conclusion .................................................................................................................. 99 6 CONCLUSIONS ........................................................................................................ 100 LIST OF REFERENCES ................................................................................................. 102 BIOGRAPHICAL SKETCH .............................................................................................. 106

PAGE 7

7 LIST OF TABLES Table page 2 -1 PLC Adaptors ......................................................................................................... 24 2 -2 Classification of video streaming ........................................................................... 30 2 -3 Quality of 25Mbps, HD video streaming ................................................................ 30 4 -1 Priority Scheme ...................................................................................................... 50 4 -2 Traffic Generation Parameter ................................................................................ 53 5 -1 D flag in NRA message .......................................................................................... 63 5 -2 H flag of ENS message .......................................................................................... 69 5 -3 A flag of END message .......................................................................................... 70 5 -4 Parameters ............................................................................................................. 78 5 -4 Parameters for analysis ......................................................................................... 89

PAGE 8

8 LIST OF FIGURE S Figure page 2 -1 DCF access scheme .............................................................................................. 17 2 -2 Beacon period structure of Home Pl ug AV ........................................................... 19 2 -3 TDMA allocation Of Home Plug AV. ...................................................................... 19 2 -4 CSMA scheme in HomePlug AV ........................................................................... 20 2 -5 Transmission sequence of UPA/DHS .................................................................... 21 2 -6 Hybrid Medium Access Protocol ............................................................................ 22 2 -7 Test floor plan ......................................................................................................... 23 2 -8 Snapshot of Gatorbytes ......................................................................................... 24 2 -9 PL C test configuration ............................................................................................ 25 2 -10 IEEE 802.11g/n test configuration ......................................................................... 25 2 -11 Link coverage versus TCP throughput .................................................................. 26 2 -12 Link coverage versus UDP throughput .................................................................. 27 2 -13 TCP throughput versus distance of direct path ..................................................... 28 2 -14 UDP throughput versus distance of direct path .................................................... 29 3 -1 Text versus graphical event analyzer .................................................................... 37 3 -2 Example of selecting desired events. .................................................................... 40 3 -3 EVGATOR configuration format ............................................................................ 41 3 -4 Generated trace log file format .............................................................................. 41 3 -5 Example of actual display window ......................................................................... 42 4 -1 MAC Scheduling scheme of Bluetooth .................................................................. 47 4 -2 Flow chart of ELRR -PP .......................................................................................... 51 4 -3 Average delay with same arrival rate .................................................................... 52 4 -4 Throughput with variant traffic load ....................................................................... 53

PAGE 9

9 4 -5 Average delay with variant traffic .......................................................................... 54 5 -1 Mobile IPv6 handover procedure ........................................................................... 58 5 -2 Mobile IPv6 handover procedure with FMDD Scheme ......................................... 60 5 -3 PMIPv6 handover procedure ................................................................................. 65 5 -4 Proposed PMIPv6 handover procedure ................................................................ 68 5 -5 IAPP MOVE notify packet format ........................................................................... 72 5 -6 The proposed PMIPv6 scheme in IEEE 802.11 network. ..................................... 73 5 -7 Network Configuration ............................................................................................ 74 5 -8 timing diagram of MIPv6 ........................................................................................ 80 5 -9 Timing diagram of PMIPv6 ..................................................................................... 81 5 -10 Timing diagram of proposed scheme1 .................................................................. 83 5 -11 Network topology ................................................................................................... 88 5 -12 Mobil e IPv6 handover latency comparison 1 ........................................................ 90 5 -13 Mobile IPv6 handover latency comparison 2 ........................................................ 91 5 -14 Mobile IPv6 handover latency comparison 3 ........................................................ 91 5 -15 PMIPv6 handover latency comparison 1 ............................................................... 92 5 -16 PMIPv6 handover latency comparison 2 ............................................................... 93 5 -17 PMIPv6 h andover latency comparison 3 ............................................................... 94 5 -18 PMIPv6 p acket loss comparison 1 ........................................................................ 95 5 -19 PMIPv6 packet loss comparison 2 ........................................................................ 95 5 -20 PMIPv6 packet loss comparison 3 ........................................................................ 96 5 -21 Packet d elay in s ession t ime ................................................................................. 96 5 -22 PMIPv6 cost comparison 1 .................................................................................... 97 5 -23 PMIPv6 cost comparison 2 .................................................................................... 98 5 -24 PMIPv6 cost comparison 3 .................................................................................... 98

PAGE 10

10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ANALYSIS AND PERFORMANCE ENHANCEMENTS IN POWER LINE, WIRELESS AND MOBILE NETWORKS By Sunguk L ee M ay 2010 Chair: Haniph A. Latchman Major: Electrical and C o mputer Engineering This dissertation focuses on the study of the performance of Powerline Communication ( PLC ), wireless and mobile network s in conveying multimedia traffic such as Internet Protocol Television ( IPTV ). We provide a comprehensive protocol performance comparison for wireless and PLC inhome network s Our result shows that the throughput of Homeplug AV is at least 18Mbps higher than that of IEEE 802.11 ( g and n ) because of improved design and the quasi -stationary channel characteristics of PLC We also sh ow that Homepl ug AV outperforms other competing PLC technologies by as much as 17.5% for TCP and 77.7% for UDP at 77% of connections and in addition achieves 100% full home coverage in our performance evaluation. We also develop ed and used in this work a network protocol visualization tool Visugator for performance evaluation of network protocols. Th is tool ha s been made freely available for public use. For distributing multimedia traffic in mobile networks a mobility management protocol is a n essential element of mobile computing including mobile IPTV. We developed a F ast Movement Detection and DAD ( Duplicat e Address Detection) (FMDD)

PAGE 11

11 scheme for Mobile IPv6 protocol as well as two improved network based Proxy Mobile IPv6 protocols over wirel ess local area network that reduc e handover latency and signaling overhead. In the Mobile IPv6 with FMDD scheme, the neighbor cache is used for DAD to reduce the handover latency. To reduce the handover latency and packet loss of Proxy Mobile IPv6, simult aneous L2 & L3 handover procedure s and a tunnel ling scheme are used. With the help of an Inter Access Point Protocol ( IAPP ) we decrease the signaling overhead of Proxy Mobile IPv6 significantly. A mathematical analysis is provided to show the benefits of our proposed schemes. Based on our analytical result s, the FMDD scheme with Mobile IPv6 can reduce the DAD latency by lookup delay in RAM from the 1 second to just several micro seconds. Our proposed Proxy mobile IPv6 protocols also can reduce up to 12.7% o f the handover latency, 29.56% of packet loss and 31.6% of signaling overhead compar ed with the conventional Proxy Mobile IPv6. The dissertation also presents a MAC scheduling scheme of Bluetooth to improve MAC efficiency.

PAGE 12

12 CHAPTER 1 INTRODUCTION Until now several technologies for Local Area Network ( LAN) have been developed and utilized for many applications. Recently the demand for high speed LAN technologies has increased greatly to support the deployment of high definition video streaming as HDTV has become more wide spread. Wireless LAN technolog y is gaining attention to provide high speed service at hotspot s in public area s like an airport and train station as a part of mobile cellular network s or independent system s using LAN technologies designed for indoor applications Home Network Many LAN technologies have been competing for preeminence as the multimedia home networking technology of choice. HDTV is usually encoded in about 25Mbps, and so to stream one or more HDTV signal requires networks with sustained data rates of more than 30Mbps at th e application layer. While traditional Ethernet networks with appropriate QoS enabled switches may be one desirable choice for this purpose as it is a mature technology that supports very high bit rates, this may not be viable solution for deployment in existing buildings due to the inconvenience in laying the Ethernet cables around the house with the associated drilling through dry walls or even concrete walls. Thus wireless and the so -called no new wires solutions become the top contenders for retrofit ting a home network in existing buildings For some time there has been strong interest in the 802.11x wireless LAN technologies as key candidates for home networking as these wireless technologies provide the ultimate convenience and mobility to users. The widely used 802.11b [1] is

PAGE 13

1 3 now considered to be less desirable for streaming HDTV because of its low data rate. However, the IEEE 802. 11g [ 2 ] provides raw data rates up to 54Mbps which may provide enough application level bandwidth to support streaming HDTV. The IEEE Draft 802.11n standard [ 3 ] has also been proposed for higher throughput and QoS enhancements. It is expected to provide raw data rates up to 600Mbps and interoperability with IEEE 802.11a/b/g. Bluetooth [4] is a wireless technology that all ows communication devices and accessories to interconnect using a short -range, low -power, inexpensive radio. While initially developed to eliminate the numerous short -range cables involved in interconnecting mobile devices (laptops, mobile phones, headsets PDAs, etc) in small networks, usually referred to as PANs (Personal Area Networks), Bluetooth has expanded in scope to encroach on wireless LANs. In the class of no new wires networks, Power Line Communication (PLC), MoCA [ 5 ] and HomePNA [ 6 ] re -use existing wires for communication and claim multimedia in home streaming capabilities for one or more channels of HDTV. MoCA uses coaxial TV cable for communication and provides raw data rate as high as 270Mbps, however the lack of available access points limits its applications. The HomePNA has similar drawback as the number of telephone jacks in a home are limited and are usually in a very inconvenient locations for installing HDTV players or recorders or the similar devices. Power Line Communication (PLC) uses the existing in home power circuit to deliver digital data and is considered to be a potentially desirable candidate for home networks as there are plenty of outlets in a home and they are in convenient locations

PAGE 14

14 In PLC, there are several competing groups The Home Plug Power Line Alliance and The Universal Power line Association (UPA) as well as Consumer Electronics Power line Communication Alliance (CEPCA). The Home plug AV [ 7 ] standard is the most recent enhancement of the 14 Mbps HomePlug 1.0 standard released by the Home Plug Power line alliance and it provides 200Mbps raw data rate explicitly targeted to support distributing data and high quality, multistream videos like HDTV and SDTV. UPA [ 8 ] proposed t he Digital Home Specification ( DHS ) with the chip -sets designed by DS2, Spain. It provides 200 Mbps data rate for AV streaming while the HD -PLC (High Definition power line communication) [ 9 ] by Panasonic, is able to support raw data rate up to 190 Mbps. Mobility Management Protocol Currently wireless local area network (WLAN) is widely spread as infrastructures for high speed wireless service not only in small indoor area but also in hot spots of public area. The series of IEEE 802.11 standard are most widely deployed WLAN technology now days. The popularity of WLAN lies on its low cost and high data rate. The IEEE 802.11 b/g [ 1 2 ] uses license -free 2.4 GHz Industrial Scientific and Medical (ISM) ratio band. T he IEEE 802.11g/ a [ 2 10 ] can support d ata rate up to 54Mbps. Since the IEEE 802.11 standard provides link layer roaming a mobile user (MN) can move to another AP without disruption of current connection within the same network. However WLAN technology cannot support seamless communication when a mobile user moves to anot her network which has a unique network address. To provide IP level mobility the Mobile Internet Protocol [1 1 ] called MIPv4/v6 was proposed by the Mobile IP working group in the Internet Engineering Task Force (IETF).

PAGE 15

15 Though MIPv6 has been developed it ca nnot fulfill the QoS requirement of real time application because of handover latency. In order to reduce the handover latency Fast handover for Mobile IPv6 (FMIPv6) [12 ] has been also proposed by IEFT. In FMIPv6 a mobile user obtains the new CoA before ac tual handover. This scheme can reduce delay of Duplicate Address Detection ( DAD) which consumes almost handover latency in the MIPv6. Also Hierarchical Mobile IPv6 (HMIPv6) [1 3 ] has been proposed to improve performance of MIPv6. This protocol minimizes the signaling cost and latency of location update of MN by adopting a local anchor point called Mobility Anchor Point (MAP). HMIPv6 is more efficient when MNs are frequently changes their point of attachment to the network. W hile several improved MIP schemes have been proposed until now MIP v6 has not spread widely in practice. MIPv6 and its enhanced schemes called host based mobility management protocol require to modify protocol stack of MN for supporting them. This modification causes incensement of complex ity and power consumption of MN. Recently a network based mobility management protocol named Proxy Mobile IPv6 (PMIPv6) [1 4 ] has been proposed by IETF NetLMM (Network based Localized Mobility Management) working group. In PMIPv6 the serving network perfor ms signaling of IP mobility management on behalf of MN. Therefore no modification of protocol stack in MN is necessary. In the PMIPv6 domain MN recognizes that it is always in home network. Though PMIPv6 is superior it also suffers from packet loss and ha ndover latency.

PAGE 16

16 CHAPTER 2 C OMPARATIVE PERFORMANCE ANALYSIS OF LOCAL AREA NETWORK TECHNOLOGIES FOR MULTIMEDIA STREAMING IN HOME Introduction A major driving force for the development of advanced high speed in local area networks especially for home is the need to support the ongoing deployment of high definition digital television at rates of about 25 Mbp s. HDTV signals may originate at one point in the home from a media center player or via satellite or cable TV I t will then be the necessary to transp ort the HDTV signal throughout the home for playing the video and audio at multiple locations or for remote recording. Streaming HDTV places very stringent Quality of Service (QoS) requirements on the underlying home networking technology in terms of delay s, jitter and packet loss probabilities This chapter presents a study of effective technologies for high speed LAN and high definition (HD) video streaming applications in a home or small office environment, without the need for installing new wires. HD videos usually are encoded at a data rate of about 25Mbps and modern LANs offer throughput of 10 100Mbps. Thus a practical home networking infrastructure should provide 30Mbps or more at the application layer. Powerline communication (PLC) and wireles s networking are among the most desirable choices of building a home network because of easy installation, with no new wires needed. The main contribution of this chapter is to analyse and conduct a real world performance assessment of the above recent wireless and PLC network technology for home and small office environments Overview of Wireless LAN and Powerline communication This section gives a brief over view of w ireless and power line communication

PAGE 17

17 technologies considered in the rest of the paper. IEEE 802.11x Wireless LAN There are two accessing methods in IEEE 802.11x technology [1,2] o ne is DCF (Distributed Coordination Function) and the other is PCF ( Point Coordination Function). The infrastructure network works with DCF and PCF while ad hoc network uses DCF. The principal approach of DCF is Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA). The basic DCF Access scheme is shown in F igure 2 1 Figure 2 1 DCF access scheme To avoid collision, physical carrier sense and virtual carrier sense are used and the physical layer uses these techniques to sense whether the medium is busy or idle. Whenever a machine decides to send a frame, a duration is specified in the duration field of the frame to indicate the ending time of the transmission. O ther machines that receive the message update their local NAV (Network Allocation Vector) thus avoiding the possible collision resulting from more than one node sending frames at the same time The IEEE 802.11g [2] is operat ed at 2.4GHz band using O rthogonal F requency D ivision M ultiplexing (OFDM ) The use of the same frequency band makes 802. 11g

PAGE 18

18 interoperability with IEEE 802.11b while providing a compatible data rate with IEEE 802.11a which use s 5 GHz band. The default minimum contention window size i n IEEE 802.11b [ 1] is 31 slots and the slot time is 20s. IEEE802.11g for compatibility, use s the same parameters. However, when no IEEE 802.11b stations are present in the network, IEEE 802.11g will use 15 slots as the minimum window size and 9 s for slot time to provide lower protocol overhead and higher throughput. The IEEE802.11n [3] uses OFDM -MIMO (Multi Input Multi Output) technology and support s not only 20 MHz channel like IEEE 802.11a/b/g standards but also 40 MHz channel s in physical layer. The performance at MAC layer of IEEE 802.11 n is enhanced by aggregating several frames into one transmission frame. The frame aggregation can decrease the transmission time of preamble and frame header under ideal channel conditions. However, the corruption of aggregation frame will reduce the MAC efficiency as the long channel time is wasted Power Line Communication The Home plug AV [7] operates in the 2 -28MHz band and it uses windowed OFDM and Turbo Convolution Code ( TCC) in the physical layer. The MAC layer uses a mix of TDMA and CSMA and supports QoS (Quality of Service) There is a Central Coordinator (CCo) in e ach of the PLC network s. The CCo broadcast s a periodical beacon frame with information of TDMA and CSMA allocations at the beginning of each beacon period This beacon period is synchronized with the AC line cycle. Figure 2 2 shows the Home Plug AV s beacon period structure. The TDMA allocation is for applications that demand QoS while t he CSMA allocation is used by applications that do not have strict QoS requirement s The

PAGE 19

19 allocations of TDMA slots in Home Plug AV is dynamic. Figure 2-3 shows TDMA allocation of Home Plug AV. Beacon Region TDMA Allocation CSMA Allocation Dyanmic TDMA Alloc Session # 1 Dyanmic TDMA Alloc Session # 2 Beacon Period Figure 2 2 Beacon period structure of Home Plug AV TDMA Session # 1 CSMA Beacon Payload SACK RTS CTS SOF RTS CTS Gap ( 100 us ) CTS SOF Gap ( 100 us ) Response Gap ( 100 us ) Only used under hidden node conditions Beacon Beacon Period Figure 2 3 TDMA allocation Of Home Plug AV. The CSMA access scheme in Home plug AV is the same as Home Plug 1.0 [15] which is a modified CSMA/CA protocol with priority that allows Home Plug 1.0 stations to communicate with each other without any centralized coordination. The space between the last frame and the incoming frame is cal led CIFS (Contention Window

PAGE 20

20 Inter -Frame Spacing) while the space between the last frame and the response is RIFS (Response Inter Frame Spacing) Home P lug 1.0 provides four priority classes CA3 CA2, CA1 and CA0. The Priority Resolution Period (PRP) in f ro nt of contention window allows higher priority packets to enter the contention window while blocking low priority packets. For the lower two priority (CA1 and CA0) classes the back off window size is 8 -16 -32 64 slots, otherwise back off window size is 8161632 slots for t he two higher priority (CA2 and CA3) classes. Figure 2 -4 shows CSMA scheme in HomePlug AV. TDMA Session # 1 CSMA Beacon Payload SACK RTS CTS SOF RTS-CTS Gap (100us) CTS-SOF Gap (100us) Only used under hidden node conditions SACK Contention IFS (100us) Priority Resolution Slots Contention Resolution Slot (35.84us) Contention Window SACK for previous transmission Beacon Period Beacon Response Gap (100us) Figure 2 4 CSMA scheme in HomePlug AV The UPA DHS ( Digital Home Specification) [8] uses OFDM modulation technique and is operat ing in the frequency band of 2-30MHz. The medium Access Control (MAC) Layer of DS2/UPA uses an Advanced Dynamic Time Division Multiplexing (ADTDM)

PAGE 21

21 scheme. This ADTDM scheme is essentially a centralized MAC protocol and supports collision free access to PLC channel to all nodes in networks There is only master node in the network and this node controls channel access of all other nodes by using a token packet. The Repeater can be placed in network to forward the sign al if the signal from master node cannot reach all slave nodes T he slave nodes does not have the right to send any packets unless it is granted this right by the master node. Only one node has the right to transmit at any moment, however, the possibility of duplicate tokens and token loss is not avoidable. Figure 2 -5 illustrates this transmission sequence of UPA/DHS. Node A (Master Node) Node B Node C Node D Token for B Token for C Token for D Token for A Token for A Token for A RX/TX Switch Time : TokenTime Figure 2 5 Transmission sequence of UPA/DHS In Figure 2 5 node A starts the transmission process b y sending data bursts to its destination and gives the right to control the channel to node B by passing the Token packet. After receiving the token from node A, node B starts transmission. After finishing data transmission, node B returns the token to nod e A. The node A gives the token to node C and D in order. Every slave node returns the token to node A after finishing data

PAGE 22

22 transmission. If a node which has a right to control the channel does not have data to send, this node immediately returns the token to node A. The HD -PLC (High definition PLC) uses wavelet transform -based OFDM (orthogonal Frequency Division Multiplexing) in the frequency band 4-28 MHz. This wavelet OFDM uses 512 real carriers with symbol lengths of 8.192s and the carriers are modul ated with PAM (Pulse Amplitude Modulation). HDPLC devices can provide maximum 190 Mbps of physical throughput or more without notches. The MAC scheme of HD -PLC [9] uses a hybrid medium access protocol composed of Time Division Multiple Access (TDMA) and Carrier Sense Multiple Access (CSMA) in each b eacon cycle. The architecture of the HD PLC hybrid MAC is shown at Figure 2 6 TDMA CSMA Beacon 1 Cycle Figure 2 6 Hybrid Medium Access Protocol In the TDMA region, a centralized channel access management is exe cuted by a central coordination node. The bandwidth schedule is announced to each node by a beacon frame. Distributed channel access management is executed in the CSMA region. In this region bandwidth is not guaranteed and the priority for each link in th e CSMA region is determined from the TOS (Type of Service) field in the IP header. If a link has a high priority then a node has short Maximum Back off time. Experiment Setup The experiment was conducted in a 2200 square feet house in Gainesville, Florida.

PAGE 23

23 The objective of this test was to determine the performance of recent PLC and wireless technologies in a real world environment. The interior walls of this house are made of wood and drywall. This can give a significant advantage for wireless technologys performance comparing to the same setup in a bricked or concrete wall environment. Figure 2 7 shows the floor plan and the testing locations of this house. Figure 2 7 Test floor plan In this experiment, we used one desktop computer as a receiver and one laptop computer as a transmitter The d esktop computer had a 3.0GHz Pentium IV processor with 512 M B Ram and windows XP professional SP2. The l aptop was equipped with a 1.87GHz Pentium IV processor, 512MB RAM and running on top of windows XP professional SP2. The software used for this test was Gatorbyte s a network throughput measurement software developed by the University of Florida and Charleston Southern

PAGE 24

24 University [ 16 ] This program is able to provide real time data visualization that shows the network status in real time. A s napshot of the Gatorbyte output screen is shown in Figure 2 8. Figure 2 8 Snapshot of Gatorbyte s Both TCP and UDP protocols are used in this t est t o evaluate the performance of IEEE 802.11g/n and PLC technology. The TCP tests used 64240 byte segments while 1472 bytes packets were sent in the UDP tests. Each test lasts for 60 seconds. This test gives us an idea of the optimal throughput of each t echnology. The setup for the PLC tests are depicted in Figure 2 -9 while the PLC adaptors used in the test are described in Table 2 1 Table 2 1. PLC Adaptors Model Name Maker Chipset Technology RD 6300 Intellon Intellon Home Plug AV HDX101 Netgear DS 2 UPA DHS BL PA100 Panasonic Panasonic HDPLC

PAGE 25

25 HDX101 and BL-PA100 power line adapter used were the over the shelf products while RD 6300 Homeplug AV power line adaptor is a reference design provided by Intellon corporation. Figure 2 9 PLC test configuration An Ad-Hoc configuration was used in the wireless network experiments. The configuration is shown in Figure 2 -10 Figure 2 10 IEEE 802.11g/n test configuration The equipment used in the wireless experiments was Linksys, WUSB300N. Both of the receivers and transmitters antenna were placed randomly and the manufacturers default settings were used. Test of UDP and TCP Protocol The experiments were conducted in 18 paths with TCP and UDP protocols. All PLC and IEEE 802.11g/n technologies have full whole house coverage in this test meaning that there was some connection along every path tested. The Netgears PLC adapter, HDX101 we used had the amateur radio tones turned on while the other PLC

PAGE 26

26 adapters did not use the amateur radio bands. In the results presented, the HDX101 data rates are reduced by 20 % to correct for the amateur radio bands. Figure 2 11 Link coverage versus T CP throughput Figure 2 11 shows the percentage of links that exceed the TCP throughput value noted at the X axis. The experiment shows that Intellons PLC adapter RD 6300 has the highest M aximum TCP performance. Figure 2 11 shows that around 77% of connections operated at more than 47 Mbps. For the Netgears PLC adaptor, HDX101, 77% of the connections operated above 40 Mbps. For IEEE 802.11n ad hoc mode and Panasonics PLC adapter BL-PA100, 77% of connection operated above 25 and 26 Mbps respectively. IEEE 802.11g ad hoc mode has lower TCP performance compared with other products but gave the most stable output for wireless TCP performance. 15 20 25 30 35 40 45 50 55 60 65 0 10 20 30 40 50 60 70 80 90 100 Throughput (Mbps)Link Coverage(%) Intellon Netgaer Panasonic 802.11g 802.11n

PAGE 27

27 Figure 2 12 Link coverage versus UDP throughput Figure 2 12 shows the percentage of links that exceed the UDP throughput value indicated on the x axis. The Intellons PLC adapter RD 6300 has the highest maximum and minimum UDP throughput. The minimum throughput of RD 6300 observed in the experiment was 57.65 Mbps and 77% of connections operated above 64 Mbps. Notably, this product had highest UDP throughput at all paths. For Netgears PLC adaptor, HDX101, 77% of connections operated above 36 Mbps for UDP. But our experiment revealed this product is vulnerable to be affected by interfer ence. As shown in Figure 2 12 the minimum throughput of HDX101 is 25.99 Mbps which is an extremely low value compared with its maximum throughput. For the Panasonic PLC adapter BL-PA100 and IEEE 802.11n in ad hoc mode, 77% of connections operated above 36 and 42Mbps respectively. This experiment shows Maximum UDP 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90 100 Throughput (Mbps)Link Coverage(%) Intellon Netgaer Panasonic 802.11g 802.11n

PAGE 28

28 throughput of BL -PA100 is higher than that of IEEE 802.11n ad hoc mode but IEEE 802.11n has higher minimum throughput than that of Panasonics PLC adaptor BL PA100. IEEE802.11g ad hoc mode has lowest throughput as we know but again gave the most stable throughput. The deviation of maximum and minimum UDP throughput is just 1.07 Mbps. In this experiment the IEEE802.11g/n standards were operated as ad hoc mode. We can expect that the throughput would be reduced to below half of our test values for IEEE 802.11g/n operating in infrastructure mode which is normally used in home and small area networks. Figure 2 13 TCP throughput versus distance of direct path The scatter plot of the TCP and UDP throughput as a function of distance is shown in Figure s 2 13 and 2 1 4 Normally wireless communication has direct relation between distance and performance. But in this experiment we could not find any direct 15 20 25 30 35 40 45 50 55 60 65 20 30 40 50 60 70 80 90 Throughput (Mbps)Distance (ft) Intellon Netgaer Panasonic 802.11g 802.11n

PAGE 29

29 relationship between distance and throughput. Especially IEEE 802.11g has almost the same performance irrespective of path distance. The performances of PLC systems have no correlation with the line of sight distance measure in this experiment. This is because PLC signals are transmitted through the convoluted power network cables whi ch reach the destination. Figure 2 14 UDP throughput versus distance of direct path HDTV S treaming T est To assess the feasibility of the distribution of a single HDTV stream in the home, a 25Mbps HDTV stream was transmitted and received using each of the technologies studied. In this experiment we made a connection between the laptop and the desktop computer using PLC and IEEE 802.11g/n technologies and the HDTV stream was played over the power line and wireless network using the PowerDVD multimedia player. 15 20 25 30 35 40 45 50 55 60 65 15 20 25 30 35 40 45 50 55 60 65 Throughput (Mbps)Distance (ft) Intellon Netgaer Panasonic 802.11g 802.11n

PAGE 30

30 A momentary video freeze phenomenon during playback occurred if the overall bit rate of the videofile w as close to or exceeded the capacity the path for the selected technology. We conducted this evaluation on 5 paths in the same house. Table 2 -3 shows the quality evolution for HDTV streaming. We evaluated the quality of video streaming in 4 categories of an informal Mean Opinion Score (MOS): 4=Very Good, 3=Good, 2=Poor and 1-Very Poor. Table 2-2 shows the classifications used. Table 2 2 Classification of video streaming Classification Playback condition 4 Very Good: Video streaming is smooth and no packet drop is observed. Or it is hard to realize the packet drop 3 Good: video stream ing has slight discontinuity or delay 2 Poor: Video streaming has evident discontinuity or delay but delay is not significant 1 Very Poor: video streaming has serious discontinuity and delay so quality of video is not tolerable. Table 2 3. Quality of 25Mbps, HD video streaming Product Path 1 Path2 Path3 Path4 Path5 Avg. Intellon RD6300 3 4 2 2 3 2.8 Netgear HDX101 2 1 2 3 2 2 Panasonic BLPA100 1 2 1 1 1 1.2 Linksys 802.11G 1 1 1 1 1 1 Linksys 802.11N 1 1 1 1 1 1 In the experiment with MPEG -2 file with a bit rate of 25Mbps, we observed a serious video freeze and go (halting) phenomenon with IEEE 802.11g and n technology. The delay of IEEE 802.11g is more serious than that of IEEE802.11n. Panasonics BL PA100 PLC adaptor also shows bad video quality. Intellons RD6300 and Netgers HDX101 PLC adaptor shows better perf ormance than other technologies as we expected. But these could not support a 25 Mbps HD video signal successfully at all

PAGE 31

31 paths. (Note: the Negrear HDX101 adapters had the amateur bands turned on for these tests). Conclusion This study presents a comprehensive evaluation of the performance of recent wireless and PLC technolog ies for home and small office environment s The throughput of IEEE 802.11 g/n and High speed PLC technologies in a real world environment was measured. The experiment s were conducted along 18 paths with TCP and UDP protocols in 2200 square feet house. HDTV streaming was also evaluated over five paths. All power line and wireless technologies tested in our experiment provide full connectivity with each other. Our overall e xperimental result s shows the Homeplug AV PLC adapter outperforms other PLC products and the IEEE802.11n/g for UDP and TCP throughput and HDTV streaming performance. The work reported in this chapter has been published in the Proceedings of 12th World Multi -Conference on Systemics, Cybernetics and Informatics (WMSCI2008) [33 ].

PAGE 32

32 CHAPTER 3 A VISUALIZATION TOOL FOR NETWORKING PROTO COL ANALYSIS Introduction Recent decades have seen exponential increase in the amount of digital information flowing across data networks. Digital information goes through several network layers before reaching the intended destination. Each network layer deals with a specific facet of the networking problem. For example, the IP layer is responsible for proper routing wh ile the physical layer deals with the transmission and reception of signals over the medium. Network layers implement a set of protocols or behavior rules that enable them to perform their intended function. Finite State Machines (FSMs) are widely used in modeling and implementation of networking protocols. Typically these FSMs are augmented by means of variables and state transition dependencies based on their values. For example, the modeling of the widely used TCP protocol will include such states as t he Slow Start, Congestion Avoidance and variables such as the duplicate acknowledgments counter used for modeling the fast retransmission and fast recovery protocol. The current generation of networking protocols and hence the FSMs that are used to implement them can are quite complex. Analytical tractability of their behavior is limited to a very small subset of their range of behavior. Network engineers typically use simulations for developing and enhancing the protocols. These simulations output various statistics that can be used to estimate and improve system performance. Several types of statistics can be collected from a simulation based on the parameters of interest. Aggregate statistics provide information over the whole duration of the simulation and hence describe the average system behavior. Dynamic behavior of the

PAGE 33

33 system can be captured by means of windowed statistics (i.e., statistics generated over a widow of time) or by means of the detailed trace file. Special events can be captured by means of log files containing the relevant event information such as the time of occurrence, event type, outcome, etc. Typical system/simulation performance verification is accomplished by comparing the system behavior to the expected behavior. Recently, there h as also been an increasing trend to support Quality of Service (QoS) for multimedia streaming both within the Small Office/ Home Office (SOHO) environment as well as more generally over the Interne. There has also been an increasing need for designing intelligent protocols that automatically adjust to changes in network conditions based on cross layer information. Networking protocols required to support such functionality are significantly more complex than the current generation of protocols. The design a nd verification of such protocols can leverage the statistics gathered from the simulations. However, in several instances it becomes important to go through a long trace file to understand the behavior of the system under certain events. One such example is the design of contention based medium access control protocols like the CSMA/CA protocol used in IEEE 802.11 Distributed Coordinated Function and in the Powerline HomePlug 1.0 MAC. In CSMA/CA networks, multiple stations use a combination of carrier sens ing and back off algorithm to obtain medium access. Collisions are common in such networks and optimal performance is typically obtained by balancing the collision costs with the time wasted due to idle contention slots. To verify the proper operation of s uch protocol, one has to go through in detail the

PAGE 34

34 trace files of the states of various stations as a function of time, including their internal variables and check if they behave as expected. However, going through a long trace of data is cumbersome and er ror prone. To overcome this problem we developed a new Visual Protocol Analyzer named EVGATOR [1 7 ] that can juxtapose and plot states of multiple FSMs as a function of time. This utility is configurable so that each state of the FSM is represented by a specific color. The EVGATOR tool proved to be of great advantage in analyzing the new network high spe ed Powerline protocols. Requirements In a generic network protocol simulation, there are multiple instantiations of nodes with each node modeled using a set of FSMs. A subset of state traces of these FSMs at each node can be of interest. Assume there are j nodes on the network and that each node has k FSMs represented as FSM [1]_S tate, FSM [2]_ State, FSM [ k ]_ State Each FSM contains mi states that can be denoted by FSM [ i ][1]_ State, FSM [ i ][2]_ State, FSM [ i ][ mi]_ State where i represents the ith FSM. At any time, each FSM [ i ] State contains only one of the mi available states of the ith FSM. Thus, the general format of the trace file information is, Time,Node[1]_ States Vec,,Node [ j ]_ States Vec where, Node[ j ] States Vec is a state vector containing the state of each of the k FSMs at Node j

PAGE 35

35 Since the set of states for a given FSM is simulation dependent, the FSM state plotter should be easily configurable to any set of states. Further, the number of Nodes, j and the number of FSMs per Node, k and the number of states of each FSM, mi should also be easily to conf igure. Apart from these, the following features will be useful to have in the FSM state Plotter, Ability to choose the color/texture by which each of the states is displayed, Ability to dynamically turn on and off the states that need to be displayed, Ability to change the size of time window that is displayed and easy means of zooming in and out of the region of interest, Ability to jump to a certain point in time, and Ability to print and copy the display region. EVGator has all the above features. Note that in practice, depending on the way the simulation is designed, it may not be possible to generate a single trace file containing data from all the FSMs. For example, each instantiation of an FSM may independently generate its trace file. In such cases, it would be necessary to preprocess the respective data files to generate a single file containing all the data which could then be analyzed by EVG ATOR Design and Imp lementation The EVG ATOR tool was developed using Microsoft Visual Studio .Net platform. The programming language we used was Visual C++, largely due to its wide acceptance in industry and academia. The EVGATOR is an event driven program and thus it reacts only when user commands or system events trigger the associate functions. For example, EVGATOR remains idle until the OS sends a mouse event to it.

PAGE 36

36 The mouse event is then recognized by the gesture recognition function, which then triggers a window re draw event to the OS to reflect the corresponding mouse gesture. The OS sends mouse events only when the mouse cursor is in the E VGATOR client window and the EVGATOR is the current active window. Design of the EV G ator User Interface Humans are more sensitive and responsive to graphical data than to text. Especially when analyzing complex situations, people usually use drawings to help themselves to reason. For example, Figure 3 1 shows two modes of the same event transition in a simul ation output. People can easily understand the state transition displayed as a rather than the textual description of the same event shown on the f igure. This comparison gives us a clear indication of the need of a visual protocol analysis tool EVGATOR. To meet the requirements, EVGATOR contains the following features: Ease and efficiency in communicating with the analyzer. T he manner in which an operator interacts with the analyzer dictates the efficiency and ease of analyzing a simulation result. We c hoose to use mouse cursor as the major User Interface (UI) for the following reasons. First, using the mouse cursor is an intuitive way of interacting with modern operating systems and people are already familiar with mouse operation. Secondly, using mouse is more effective if the User Interface (UI) is reasonably designed However, we did not completely abandon the usage of a keyboard. It is undeniable that under some circumstances the keyboard is more efficient than mouse operations. For example, to jump to a specific time stamp (if the value of the time stamp is known)

PAGE 37

37 the operator simply needs to type the value into an edit box and that would be faster than using mouse to browse to the desired time mark. Ease to range magnification and contraction (Zoom). Figure 3 1 Text versus graphical event analyzer Some events rarely happen except when simulated for a certain amount of time. Thus usually protocol designers will often simulate a scenario for a long time in order to

PAGE 38

38 find out the possible design flaws. In this case, the time line in the analyzer could become quite lengthy. Thus an analyzer should be able to zoom into a specific time frame to reveal detailed protocol state transition in that time range. It also should have t he ability to zoom out to display a global view of the state transitions to find out the correlations between events. To achieve this goal, we designed mouse gestures to represent zoom in and out. Designers can use the mouse to mark a range of time which r epresents the time frame for operation. If designers click within the range, then the analyzer assumes it is a command of Zoom In. Otherwise it will zoom out so that the marked time frame occupies half of the window. Designers can easily mark various len gths of time frames to achieve various zoom factors. If designers want a specific zoom factor, they can also use the keyboard to type the value into the zoom factor edit box. To prevent ambiguity, the marked time frame is also shown as a text display on th e display window. Ease of marking a specific time. When examining simulation results, designers usually have to look at various time stamps to understand the causes of aberrant events. Without the mark function, it would be difficult to move back and forth between time frames. In the EVGATOR system, designers can achieve this effect by simply clicking on the display window. To help designers understand the marked time, the EVGATOR also shows the marked time value in the display window. Ease of navigati on throughout the whole time line.

PAGE 39

39 Since the simulation results are usually large, a display window width is not able to display all events at a time. To improve the quality of the event browsing experience, we use scroll bars to travel between time frame s. Since scroll bars are elastic to the mouse movements, designers can drastically move the mouse to achieve large step jumps or click on the end arrow box to achieve small scale time line adjustments, either forward or backward. Flexibility in changing both event display colors and event names. Since humans are sensitive to colors, EVGATOR displays different events with different colors. Meanwhile, different protocol designs may have different event naming and thus the EVGATOR is elastic enough to adapt to these minor changes without recompiling the source code. To achieve the above goals, E VGATOR reads a configuration file before it starts to parse log files. Designers can modify the configuration file to suit their needs. For example, if a protocol has a new state called Preparation, designers can open the configuration file using a text editor, and add a new line of the event name and the associated representing color. The EVGATOR will automatically add the new event into its control panel and in the representing diagram. The E VGATOR is not only elastic to the offline event representing modifications, but it is also adaptive to real time changes as well. For example, if designers are looking for a specific event, they can turn off other events to make the desired event stand out in the final diagram as shown in the Figure 3 -2. The above mentioned EVGATOR features makes using EVGATOR for analyzing protocol events efficient and easy. It can also significantly reduce the time required to debug a simulator.

PAGE 40

40 Figure 3 2 Example of selecting desired events. Configuring the EVGATOR As we mentioned earlier, the EVGATOR requires a configuration file for state presentations. The example is shown in Figure 3 3 The first line in shows a state name: INITIALIZATION and the second line shows the color that represents this event. Each statement occupies one line. E VGATOR supports 100 different states and 255 nodes. Users do not have to specify the total number of states in the configuration file. EVGATOR will determine the effective state names and number of states. Trace File Format To translate text log files into diagram forms, EVGATOR has to parse the recorded information. The sample l og file format is shown in Fi gure 3 -4.

PAGE 41

41 Figure 3 3 EVGATOR configuration format Figure 3 4 Generated trace log file format The tokens are defined in the configuration file. When EVGATOR sees a text string that is equal to a token, the index of the token is stored into an internal state queue associated with the starting time stamps. If a text is not recognizable during parsing, the EVGATOR will skip that text. The EVGATOR will keep parsing every text

PAGE 42

42 string in the log file until the end of the file. When a token is recognized, the ending time stamp of the previous state is also decided. After a whole file is parsed, the EVGATOR will start to display each state with the colors define d in the configuration file. Since the window height and width is limited and is usually smaller than the simulation duration, the real data displayed is a small portion of the w hole data as shown in the Fig. 3 5 To help designers browse, a Zoom In and Zoom Out command is required. Zoom In/Out To implement Zoom In and Zoom Out, we need to calculate the Zoom Factor. Initially the Zoom Factor is defined as one. Before calculating the zoom factor, the EVGATOR first translates the marked time frame to coordinates of the display. The translation can be done by Eq. 3. 1 M 1 = ( Mark_ Time1 Min_ time ) Zoom_ Factor M 2 = ( Mark_ Time2 Min_ time ) Zoom_ Factor ( 3.1) Figure 3 5 Example of actual display window If the mouse gesture is a Zoom In command, that is, the user makes a mouse click within the marked time frame, the zoom factor can be calculated by dividing the current display time frame with the marked time frame as sh own in Eq. 3. 2.

PAGE 43

43 21 12x-x Zoom_Factor = M-M (3.2 ) If the command is a Zoom Out, the zoom factor is calculated by dividing the marked time frame with the current display time frame as shown in Eq. 3. 3. 12 21M-M Zoom_Factor = x-x (3 .3 ) Since the simulation log file is huge, we need a proper data structure to represent state transitions. An event queue is used in EVGATOR. The queue is dynamically allocated and deallocated in order to minimize the system memory requirement. To reduce window flickering due to the updating of display contents, we record the areas that are affected by the current operation and update those areas only. This wa y, only a small portion of the display areas needs to be updated at a time. Experience The EVGATOR utility was very useful to us in developing the Medium Access Control protocol of HomePlug AV, which is intended to be the successor to the HomePlug 1.0 Standard [ 1 5 ]. During the development and analysis, we made extensive use of the EVGATOR functions described above. EVGATOR was particularly useful in the verification of the simulation and in understanding some of the rare/unexpected events that occurred. In the former case, the visual plot of the MAC and Physical Layer state as a function of time proved to be a valuable tool in trying to find any erroneous behavior. In the later case, when an unexpected and very rare event happened, the GUI helped us to q uickly go through the set of events that caused the final event. For example, in the process of designing the CSMA/CA protocol, we found that searching for the causes of nodes moving to the

PAGE 44

44 back off state using the text based log file is painful since ther e are so many different state transitions occurring at the same time. However, using EVGATOR we can easily find the cause using the displayed diagram. In Figure 3 -2 we found the cause of the back off is because of the RTS collisions. Conclusion With the h elp of the EVGATOR, we significantly reduced the protocol design process and the time spent on debugging the simulation results. Designers who work on FSM based simulators can also benefit from the visualized expressions of state transitions. Though EVGATO R is helpful, it cannot directly parse trace logs from other simulators like NS -2 or Opnet. However, users can readily transform the given log formats to the format EVGATOR accepts. We also expect the inputs from industry and academia to help us keeps developing and improving EVGATOR to meet majority of users needs, possibly with automatic data format conversion of the most popular simulation environments. EVGATOR is now available on UF LIST website (http://www.list.ufl.edu/VisuGATOR ) as a public domain freeware, with some required rights held by the University of Florida to the present source code. The work reported in this chapter has been published in the International Journal of Software Engineering and Its Applications [17].

PAGE 45

45 CHAPTER 4 ENHANCED LIMITED ROUND ROBIN MECHANISM U SING PRIORITY POLICY OVER BLUETOOTH NETWO RK Introduction The smallest Bluetooth unit is called a piconet, which is composed of one master node and several slave nodes (up to seven).In a piconet a Bluetooth device can be operate in active mode or sleep mode (Sniff, Hold and park mode ). However just one master and seven active mode slaves are allowed in a piconet. If some nodes desire to switch the state to active mode, some active nodes should change the stat e from active mode to sleep mode when seven slaves are already operated in active mode. All nodes in a same piconet should follow same frequency hopping pattern. Multiple piconets can also exist in the same area and be connected via a bridge node, creatin g a scatternet. In a Bluetooth system, full -duplex transmission is supported using a master -driven TDD (Time Division Duplex) scheme to segment the channel into 625s time slots, which are alternatively distributed between the master and the slaves. As such, the master sends a poll or data packet to a slave using the even -numbered slots, then the slave sends a packet to the master using an oddnumbered slot immediately after receiving a packet from the master [ 4 ]. Thus, the MAC scheduling in Bluetooth is c ontrolled by the master and executed using master -slave pairs. All Bluetooth devices in a same piconet have to synchronize the clocking exactly and a node which has privilege of using the slot will not release the resource of transmission to other nodes ev en if it does not have any data to transmit. Hence Bluetooth seems to be inefficient but it can support more reliable connection. So Bluetooth is appropriate for high quality interconnection between mobile devices and many researchers have proposed algorit hms to enhance the performance of Bluetooth network [ 18 -22 ].

PAGE 46

46 Bluetooth systems support two types of virtual data communication link between the master and slave: a Synchronous Connection Oriented (SCO) link and Asynchronous ConnectionLess (ACL) link. An SCO connection supports a circuit oriented service with a constant bandwidth using a fixed and periodic allocation of slots An SCO connection is suitable for latency -sensitive multimedia traffic like voice traffic, whereas an ACL connection supports a pac ket oriented service between the master and slave for data. An SCO connection only uses one time slot, while an ACL connection can use one, three, or five time slots according to the data packet size [ 4 ]. The Round Robin (RR) scheme is a default MAC schedu ling algorithm for Bluetooth that uses a fixed cyclic order. The POLL packet does not have any information and just gives the polled slave the privilege of transmitting packet in next slot. If the polled slave does not have any data to transmit, it replies to the master by sending a NULL packet which also does not have any information. As a result, numerous slots can be wasted with POLL or NULL packet exchanges in the case of no data to transmit. Despite the introduction of several Bluetooth MAC scheduling algorithms to reduce these slot waste, problems still remain. Accordingly, this section introduces a simple and efficient MAC scheduling algorithm to reduce slot waste and enhance the performance of Bluetooth. M AC scheduling scheme of Bluetooth The principal MAC protocol in Bluetooth is a polling scheme that uses a master driven time division duplex (TDD). Here, a slave is only allowed to transmit a packet to the master immediately after receiving a packet from the master. Plus, the master can only use an even numbered slot to send a packet to a slave, while the slave can only reply to this packet using an oddnumbered slot. The MAC scheduling scheme for a

PAGE 47

47 Bluetooth system is shown in Figure 4 1 The TDD slots are shared by the SCO and ACL link in this f igure Figure 4 1 MAC Scheduling scheme of Bluetooth In the Pure Round Robin (PRR) scheme, every slave has the same opportunity to send one data packet even when they have no packet to transmit. Once the master polls a slave, the next time slot is then assigned to the slave without reference to whether the slave has data to transmit or not. Consequently, numerous slots can be wasted when POLL packets are sent by the master and NULL packets are sent by the slaves in the case of no data packets to transmit. Therefore, the PRR scheme is an inefficient scheduling algorithm when the traffic in the network is asymmetric and several Bluetooth MAC scheduling algorithms have already been proposed to improve the system performance [ 1822 ]. Some works assu me that the Master knows upto date information of slaves queue status [ 18 20]. Some approaches do not require this information [ 22 ]. Some researchers also proposed low power mode based scheduling policy [ 21 ]. In [ 18 ], a Master -Slave Queue-State -dependent Packet scheduling algorithm is proposed, where a free bit in the Bluetooth payload header is used by a slave to inform

PAGE 48

48 the master of the next available data. Based on this feedback bit, the master classifies all the master -slave pairs into one of four states, and a higher priority assigned to a pair that utilizes the slots more efficiently than the other pairs. However, this algorithm assumes that the master has updated information on the slave queue at all times, whi ch is not actually available given the current Bluetooth specifications as the slaves can only provide information on their queue state when they send a packet to the master. In addition, a slave starvation problem can occur for low priority pairs when hi gher priority pairs always have packets to transmit In [ 19 ], HOL -Priority Policy (HOL -PP) is proposed. This algorithm use similar priority policy with [ 18 ]. The master schedules on the basis of the head of line (HOL) packet size at the Master and slave queue. Additional two bits are required to distinguish 4 possible HOL packet size for classifying the master and slave pair into three classes by slot utilization. However this policy assumes the master has up-to -date information of slave queues status li ke [ 18 ]. In [ 20 ], the authors use the similar priority policy as in [ 18 ] and [ 19 ] to improve the performance of the system considering both throughput and delay on each Master Slave pair in scheduling decision. They use the amount of link utilization to as sign a priority class based on queue status of master and slave pairs. But they also assume that the master knows up to date information of the slave -to master queue status. Some researchers proposed MAC scheduling policy to optimize power consumption and improve slot utilization by using a low power mode in Bluetooth [ 21]. In this research authors proposed variable sniff interval and variable serving time on the

PAGE 49

49 base of slot utilization. However this scheduling policy is not adequate for dynamic traffic a nd requires additional information. In [ 22 ], several scheduling algorithms are compared. The Exhaustive Round Robin (ERR) is a simple scheduling policy that uses a fixed order like RR. Yet, since the master does not switch to the next slave until both th e master and the slave queues are empty, there is a danger that the channel can be captured by stations generating a higher traffic than the system capacity. Thus, the Limited Round Robin (LRR) was proposed to solve this capture effect Although the LRR als o has a fixed cyclic order and is exhaustive, like the ERR, parameter t is adopted to limit the number of transmissions (tokens) that can be performed by each pair per cycle. The maximum number of transmissions per cycle then limits the cycle length and avoids the capture effect. Nonetheless, despite an improved throughput and fairness when compared with the RR and ERR, the LRR still suffers from a slot wastage problem under asymmetric traffic conditions. Accordingly, on the foundation of the MAC scheduli ng algorithms in [ 18] and [ 22 ], this section introduces an Enhanced Limited Round Robin with a priority policy (ELRR PP) based on slot utilization to improve the performance. Plus, fairness is maintained using a maximum number of transmissions per cycle. ELRR PP Algorithm The ELRR -PP [23] operates in accordance with the master -slave queue status. As such, four classes of priority are assigned to the master -slave pairs based on the existence of data in the respective queues, which is similar to the method in [ 18 ], except the proposed method uses the current queue status for the master and previous queue status for the slave.

PAGE 50

50 Table 4 1. Priority Scheme Priority Queue status Class 1 Master has packets in queue and the slave sent packets to the master during previous turn Class 2 Master has packets in queue and slave did not send packets to the master during previous turn Class 3 Master has no packets in queue and slave sent packets to the master during previous turn Class 4 Master has no packets in queue and slave did not send packets to the master during previous turn Hence, additional information on the slave queue is not necessary, in contrast to [ 18 ]. This priority scheme is described in Table 4 1. Based on the classes, the slave with the highest priority is polled first. It is also assumed that class 2 has priority over class 3, as there is a possibility that the slave has no data to send to the master In the case of several pairs in the same class, the pair with the largest amount of data in the mas ter queue has priority. But for the class 2, the pair with the smallest amount of data in the master queue has priority to relegate this pair to class 4. The ELRR -PP algorithm is described in Figure 4 -2. To avoid starvation of the low priority pairs, a maximum number of transmissions per cycle is adopted for each class, where Max1 is the maximum transmission number for class 1, and Max2, which is smaller than Max1, is the maximum transmission number for cl ass 2 and class 3. Parameter t is used to count the number of transmissions. To prevent slot waste, the polling interval for class 4 pairs is modified. For the first cycle, such an inactive pair has the same opportunity to be polled. But, if it still has no packets to send in the next cycle, the polling interval for this pair is increased

PAGE 51

51 linearly by one cycle to limit the chances of transmitting packets. Parameter i is used for this purpose in Figure 4 -2 Figure 4 2 Flow chart of ELRR -PP Performance Evaluation To evaluate the proposed algorithm, a discrete event simulator was used, and all the simulations were based on a single piconet consisting of one master and six slaves,

PAGE 52

52 where the master had a corresponding queue for each slave. The t raffic mode considered was ACL. The traffic at the master and each slave was generated independently in accordance with a Poisson process. When comparing the proposed algorithm with other algorithms, the maximum number of transmissions per cycle for the LR R was set at 4, while for the proposed ELRR PP it was set at 8 for class 1(Max1=8) and 4 for class 2 and class 3 (Max2=4). Figure 4 3 shows the average delay when each master -slave pair had the same arrival rate. As shown in Figure 4 3 the ELRR -PP produced a lower delay, however, the difference compared to the other algorithms was not significant. Figure 4 3 Average delay with same arrival rate To evaluate the delay characteristics when each pair had a different traffic generation rate, the arrival rat e was changed, as shown in T able 4 2. The arrival rate of the 6th master -slave pair was changed from 0.03 to 0.27 to increase the system traffic

PAGE 53

53 from 0.4 to 0.88. As such, the arrival rate for slave 6 was increased from 0.03 to 0.27, while that for the mas ter was selected randomly between 0.03 and 0.27. Table 4 2 Traffic Generation Parameter Piconet M1 S1 M2 S2 M3 S3 Arrival rate 0.2 0.02 0.05 0.007 0.007 0.05 Piconet M4 S4 M5 S5 M6 S3 Arrival rate 0.001 0.002 0.002 0.001 Variable Variable Figure 4 4 Throughput with variant traffic load Figure 4 4 presents the throughput for the 6th master -slave pair with a variant traffic load. As Expected, the PRR produced a lower throughput than the other algorithms due to the increased traffic load for slave 6. Meanwhile, the ELRR -PP produced a higher throughput than the LRR, and higher or similar throughput to the ERR

PAGE 54

54 Figure 4 5 shows the average delay with a variant traffic load for slave 6. The delay of the PRR suddenly increased when the traffic load fo r slave 6 increased. Also, the delay of the LRR increased significantly as the traffic load increased The ELRR -PP produced a lower delay than the PRR and LRR, and comparable performance with the ERR. Figure 4 5 Average delay with variant traffic Conclusion This paper proposed an enhanced Limited Round Robin with a priority policy for data exchange within a Bluetooth piconet, and its throughput and delay performances were evaluated. The proposed algorithm utilizes the status of the master and slave to schedule the slots effectively, yet additional information on the slave is not necessary. In simulations, t he proposed ELRR -PP produced a better performance than the existing PRR and LRR scheme. In comparison between proposed ELRR -PP and conventional

PAGE 55

55 ERR, our proposed ELRR -PP has better or comparable performance when the ERR is vulnerable to a channel capture problems. The work reported in this chapter has been published in the International journal of Advanced Science and Technology [23].

PAGE 56

56 CHAPTER 5 EFFICIENT MOBILITY MANAGEMENT SCHEME OVER WIRELESS LOCAL AREA NETWORK Introduction While wireless local area network (WLAN) has been designed and utilized for small area application with limited mobility support, WLAN is gaining attention to support high speed wireless service for wide hotspot area as a part of mobile cellular network such as Universal Mobile Telecommunication system (UMTS) and General packet Radio Service (GPRS) or an independent system. The mobile Node (MN) can move around among Access Points (AP) which are in the same network without losing current connection because the IEEE 802.11 standard provides a link layer roaming. This process of link layer switch is called L2 hand over and has AP probe, authen tication and association phases in IEEE 802.11 based WLAN. However the IEEE 802.11 standard cannot support MNs continuous communication with AP at every time for public user since there are many separate networks which have a unique IP address in wider hot spot area. If the MN moves to an AP which is not in a same network it has to reconf igure its IP address again which makes current connection becomes disabled. The Mobile Internet Protocol (MIP) supports IP level mobility to provide seamless conn ectivity to MNs when it moves to another AP which is in a different subnet. The MIP v6 [ 11 ] was proposed by the mobile IP working group in Internet Engineering Task Force (IETF) to manage movement of MNs in wireless IPv6 network. In MIPv6 protocol, IP level mobility is provided by binding a Home of Address (HoA) and Careof Address (CoA) at a Hom e Agent (HA) and Correspondent Node (CN). When a MN moves to a new Access Router (AR) in different subnet it should obtain a CoA using

PAGE 57

57 Router Advertisement (RA) message from the new AR which is Foreign Agent (FA) in this case and notifies this CoA to the H A and CN to bind the CoA and HoA. This procedure of network layer switch is called a L3 handover. Packets sent to the MN cannot be delivered until the binding CoA and HoA is finished at HA. So this disruption may degrade the quality of service especially for real time and delay sensitive traffic such as a multimedia streaming and voice over IP (VoIP) service. To improve the handover performance of MIPv6 several schemes have been proposed such as Fast Handover for Mobile IPv6 [ 12 ]. To provide the MNs mobility in IPv6 network, the protocol stack of MN is required to be modified for supporting the MIPv6. This modification is too complex considering with limited capacity of the MN and participa tion of MN in mobility signaling causes increase of power consumption of the MN. The proxy Mobile IPv6 (PMIPv6) [ 14] is a network based localized mobility management (NetLMM) protocol which has been proposed by IETF NetLMM working group. In this PMIPv6 th e modification of MNs protocol stack is not necessary and MN is not involved in IP mobility related signaling. On behalf of the MN, the network is liable for managing IP mobility. The PMIPv6 guarantees the localized mobility support of MN when MN roams wi thin a local mobility domain region. Enhanced Handover Scheme of MIPv6 w ith Fast Movement Detection & DAD Overview of MobileIPv6 (MIPv6) The Mobile IPv6 (MIPv6)[ 11] is a host based mobility management protocol to support global mobility of the MN. The MN uses a permanent HoA and temporary CoA in the MIPv6 protocol. When the MN enters a foreign domain the MN should obtain the new CoA from router advertisement message. After configuration of new CoA Duplicate Address Detection (DAD) procedure is performed. Then the MN registers

PAGE 58

58 this CoA with the HA through binding update (BU) message. When the MN moves away from home network the HA works as a stationary proxy. NAR PAR HA Binding Update (New CoA) MN CN Packet L2 EstablishmentRendezvous Time Registration to HA Periodic Unsolicited Router Advertisement Router Solicitation Router Advertisement CoA Configuration and DAD (upto 1000ms) Binding Acknowledgement Packet Figure 5 1. Mobile IPv6 handover procedure The HA intercepts packets destined to the HoA of the MN and forwards these packet by tunnelling to the CoA of the MN. For efficient transmission of packets the Mobile IPv6 has a route optimization scheme. In this scheme the MN sends its new CoA to CN by BU message. After receiving this BU message the CN sends packets to the

PAGE 59

59 MN directly. However packets from CN are delivered to the MN via the HA until the CN receives the new CoA of the MN. Figure 5 -1 shows handover procedure of mobile IPv6. The MIPv6 handover procedure consists of movement detection, new CoA configuration and location update. These procedures cause a long handover latency which can not acceptable for real time multimedia application. In order to improve the handover latency the various extensions of MIPv6 such as fast handover for Mobile IPv6 (FMIPv6 ) [12] and hierarchical MIPv6 (HMIPv6) [13] have been proposed. Though a lot of enhanced MIPv6 schemes have been reported over the past year the MIPv6 has not deployed widely in practice because of heavy specification which has to be implemented at small mobile node for supporting MIPv6 Movement Detection Procedure in MIPv6 The primary aim of movement detection is to identify L3 handovers. In MIPv6 movement detection generally uses Neighbor Unreachability Detection to determine when the default router is no longer bi directionally reachable, in which case an MN must discover a new default router on a new link. However, this detection only occurs when the MN has packets to send, and in the absence of frequent router advertisements or indications from the link -layer, the MN might become unaware of an L3 handover. After a change of link layer connection the MN must detect any change at the IP layer before it can signal the change to the network. MIPv6 uses RS and RA to detect changes of IP network prefix. This is part of the standard router discovery protocol. The protocol contains built -in timers, these timers prevent a router from sending immediate responses to RS in order to prevent multiple nodes from transmitting at

PAGE 60

60 exactly the same time and to avoid long-range periodic transmissions from synchronizing with each other. These are the significant delays since they interfere with the MIPv6 movement detection algorithm thus preventing mobility signaling for up to 1000ms [ 11]. Proposed MIPv6 Handover Scheme with FMDD NAR PAR HA Binding Update (New CoA) MN NAP Stored RA Link layer establishment Registration to HA New Router Solicitation New Router Advertisement Binding Acknowledgement Packet L2 Association Response L2 Association Request RA message Lookup delay for DAD Rendezvous Time Figure 5 2 Mobile IPv6 handover procedure with FMDD Scheme Most handov er mechanisms for MIPv6 make use of movement detection algorithms, which use received RA information and configuration of new CoA. However,

PAGE 61

61 in movement detection, RS delays the transmission for a random amount of time [6]. This time serves to alleviate con gestion when many hosts start up on a link at the same time Also, RA must be delayed by a random amount time. In DAD procedure, after generation of the CoA, an MN has to check CoAs uniqueness [ 39 ] [40 ]. The duration required to complete DAD is up to 1 second. These delays for RS, RA, and DAD are inherent to MIPv6. In this section, we illustrate our proposed fast movement detection and duplicate address detection mechanism for fast handover in MIPv6. Figur e 5 -2 shows the MIPv6 handover procedure with fast movement detection & DAD (FMDD) scheme. Movement Detection using Stored RA message For movement detection, the MN performs a scan to see APs through probes. The result of the scan is a list of APs information. Authentication is performed, and then the MN sends the (Re) association request message with its MAC address. The AP grants association by sending the Association Response Message. As soon as association is made, the AP sends Stored RA (SRA) to an MN with MAC address in Association Response message [ 38 ]. However, in our proposed Movement Detection scheme, w hen the MN receives the SRA message from an AP during an L2 handover, the MN compares the prefix of the RA message with existing own prefixes in th e cache. If the prefix is different, the MN can generate a new care of address (CoA) using the stateless address auto-configuration method and the prefix option allowed in RA messages. Before finishing an L2 handover, the MN sends the RS message by unicast to the new AR using the received SRAs source address included in the Association R esponse Message. Unicast is of great significance since it can reduce the network load and avoid the Random delays attributed to RS and RA.

PAGE 62

62 An MN is able to help smooth proxy handover by adding the new CoA generated by an MN to the RS as an option to pr ovide interoperability with normal nodes. This happens by using a bit in reserved field and notifying the node that follows by the scheme that we offer. We name this bit as the New MN (NMN) and the two options as Previous MNs CoA and previous ARs global address. This modified new RS message is also named as NRS. When the new AR receives the NRS message from an MN, the AR performs DAD as well as movement detection by using the neighbor cache. DAD configuration and Lookup Algorithm The proposed DAD method i s to use of the neighbor cache of an Access Router (AR) with sufficient buffer size to allow many nodes to attach simultaneously. Firstly, the MN allocates this generated New CoA in its interface without DAD. We need to consider how a neighbor cache and th e Neighbor Unreachability Detection procedure perform DAD As stated [ 39 ], a neighbor cache contains one entry for each neighbor to which the node has recently sent messages. Each entry of a neighbor cache may be generated by the RS, the Neighbor Solicitat ion (NS) or the Unsolicited Neighbor Advertisement (NA) in the case of router. The entries generated in this way are maintained in the state of stale until traffic is sent to the neighbor. Since the neighbor cache has the list of all hosts of the link, whi ch the AR manages, we can compare the entry of the neighbor cache with an MNs MAC address, which is included in each NRS message transmitted from MN to AR. If the addresses of all nodes are generated on the link by stateless auto-configuration, we dont h ave to consider DAD [ 40]. However, if the link local address of a node is changed manually in the middle of use or generated by some exceptional process the ARs cannot maintain the entry of the node in the neighbor cache. We propose that ARs should be able to receive solicited multicast NS

PAGE 63

63 messages for normal DAD of exceptional nodes. Since the solicited NS message is sent by multicast, the ARs can receive this message by modification of its interface [ 40 ]. The lookup in the neighbor cache on N is the DAD procedure. The DAD consumes an extremely short amount of time, typically a few micro second units, such as Longest Prefix Matching speeds in routing table. The use of neighbor cache for movement detection and DAD gives an additional advantage of obtaining alternative addresses because addresses are managed in the neighbor cache. If the new AR receives the NRS message from an MN, it can check the entry of the Neighbor Cache for movement detection and DAD. If there is no duplication, an NRS message should als o be included in the Neighbor Cache. If not, it must find an alternative address. This alternative address can be chosen in the pre -con f igured table made by the router. This alternative address is inserted as a new entry in the Neighbor Cache. If this procedure is completed, the AR cans unicast the RA message to the MNs Link Local Address of the destination address. Certainly this RA can also be modified like the RS message by adding a 2bit D -flag to the Reserved flag and including the New MAC address and New link -local address as options in the option field in case of address duplication. We name this new RA message NRA. Table 5 -1 defines the D bits. When an MN receives an NRA the MN has to be operated by D bits. Table 5 1. D flag in NRA message D flag Description 00 Must change MAC address (can not apply in IEEE 802.11) 01 Can use the new CoA 10 The new CoA is allocated by other node 11 Cannot used W e denote LD as the address lookup delay, which is the time required to check an MNs MAC address for movement detection and DAD in the Patricia Trie search.

PAGE 64

64 Accordingly the address lookup delay is given as: LD = tDA x N ( 5. 1) Where tDA is the delay for access and comparison operations in RAM and N is the number of lookups in Patricia Trie. This Patricia Trie has the worst performance in line per minute. We use this algorithm in order to show the lookup time of the worst performance. Under the present circumstance, since a memory access requires fro m 60 to 100 nsec [ 41 ] and a comparison requires 10nsec in DRAM [ 42 ], we can use the value of DAt as 70 and 110nsec. In the Patricia Trie case, since lookups require accessing memory 48 times in the worst case, the N value is 48. Henc e, LD .T he calculated lookup delay is very small. Efficient PMIPv6 Handover Scheme over Wireless Local Area Network Overview of Proxy Mobile IPv6 (PMIPv6) The proxy Mobile IPv6 (PMIPv6 ) [ 14 ] is a network based mobility management protocol to supports localized mobility of the MN in PMIPv6 domain. The Mobile Access Gateway (MAG) and the Local Mobility Anchor (LMA) are main functional constituents of PMIPv6 network. The MAG performs mobility management on behalf of the MN and it detects the movement of MN and transmits the binding update message to the LMA. The LMA which has functional capabilities of HA as defined in MIPv6 protocol is the topological anchor point for the MNs home network pre fix in the PMIPv6 domain. It is responsible for managing the MNs reachability state. The PMIPv6 supports mobility of the MN which does not have MIP function in its protocol stack. From the help of the MAG and LMN the MN recognizes it is always in the home network whenever the MN roams in the same PMIPv6 domain. Figure 5 3 shows a

PAGE 65

65 PMIPv6 handover process of the MN. NMAG PMAG AAA LMA AAA request With MN-ID AAA Reply with profile Proxy Binding Update Proxy Binding Ack (Home network Prefix)Router Advertisement MN Tunnel CN Packet MN Attachment Packet (MN-ID, Home Network Prefix ) (MN-ID) L2 Establishment Rendezvous Time Registration to LMA [Proxy-CoA:LMA][MN-HoA: CN]Data [MN-HoA:CN]DataData Packet Data Packet Tunneled Data Packet [MN-HoA:CN]Data NMAG obtains MN-ID from MN Attachment Event Packets will be lost until registration Figure 5 3 PMIPv6 handover procedure Once the MN enters a region of new MAG (NMAG) the link layer between the MN and the NMAG is established. The MN sends MN -identifier to the NMAG for authentication. Using this identifier the NMAG obtains the MNs policy profile from policy store such as an AAA (Authentication Authorization Accounting) sever. This policy profile has essential parameter like as home network prefix and address

PAGE 66

66 configuration mode for providing network based mobility services. The NMAG sends AAA request message to the AAA server then this server performs MN authentication and sends the MNs profile to the MAG. The NMAG sends Proxy Binding Update (PBU) message to the LMA for updating the location of the MN. On receiving the PBU message the LMN establishes a binding between MNs home network prefix and address of MAGs egress interface which is called Proxy care of Address (Proxy CoA). The LMA sends Proxy Binding Acknowledgement (PBA) message with the MNs home network prefix to the NMAG as a response of the PBU message. By exchange of these messages bidirectional tunnel between the NMAG and the LMA is established. Then the NMAG send Router Advertisement (RA) message which has MNs home network prefix. In the same PMIPv6 domain the MN receives same home network prefix by RA message. So the MN can recognize it stays at the home network. Proposed PMIPv6 pr otocol 1 In this section we propose an efficient network based handover scheme on IEEE 802.11 network without changing mobility stack in IEEE802.11 based Mobile node for reducing handover latency The MN does not need to transmit or receive any management packet to support IP mobility. The same terminology with PMIPv6 is used and MNs localized mobility is considered. Our assumption of this proposed scheme is that AP can send information of the MN to the MAG directly and place some information in a management packet like a beacon signal Also we introduce a new signal between NMAG and PMAG. L2 Handover Procedure The lin k layer handover of IEEE 802.11standard has typically three distinct phases: discovery, authentication and reassociation [ 24] The MN scans wireless medium to

PAGE 67

67 obtain information of APs. From this information the MN selects a candidate AP to establish a new radio link base on the strength of received signal from APs. The MN can broadcast probe request message to get probe reply from AP or wait passively the periodic beacon signal from AP for this purpose. In this proposed scheme a MAG notifies its address periodically to own APs which are connected directly with the MAG. These APs have a distinct buffer only for an address of the MAG. The AP stores recently received MAGs address periodically in this buffer and puts this address into flexible frame body of periodic beacon message. A MN measures the strength of signal from neighboring APs and initiates a L2 handover when the strength of beacon signal from the new AP is bigger than that of current AP. From the beacon message the MN knows an address of MAG connected with new AP. If the address of MAG connected with new AP is not changed comparing with that of old AP, the MN needs to do just a L2 handover. Otherwise the MN has to do not only a L2 handover but also a L3 handover. After authentication is made the MN sends reassociation req uest message to new AP with MNs ID which is a MAC address in IEEE802.11 WLAN and previous AP (PAP)s address. The AP grants association by transmitting the reassociation response message. L3 Handover Procedure In our proposed handover scheme, L3 hand over procedure is performed with L2 handover simultaneously. Figure 5 -4 shows proposed handover procedure. After discover and authentication phase the MN sends reassociation request message to a new AP (NAP) and the NAP delivers this packet to a New MAG (NMAG) to notify attachment of a new mobile node and its information like a MN -identifier and the PAP address. Then the NMAG requests MNs profile by sending AAA request

PAGE 68

68 message to the policy store like AAA server. The MN -Identifier is used for this proces s. NAP PAP NMAG PMAG AAA LMA Beacon signal w/ NMAG address Reassociation request (MN-ID, PAP address) AAA request AAA Reply Enhanced Neighbor solicitation (ENS) Tunnel M-PBU PBA SPR message Packet Forward Pckets Packets via the PMAG New Packets via the LMA and the NMAG Enhanced Neighbor Advertisement (ENA) Router Advertisement (Home network Prefix) Reassociation Reply MN Tunnel CN Packet MN measures signal strength of APs. Handover initiates when NAPpower > PAP powerNMAG address Link layer establishment Rendezvous Time Registration PMAG start storing packets to MN in its buffer after receiving ENS message from NMAG Packet buffering periode AP stores MAGs address in a buffer and puts this address at periodic beacon signal (Multicast to the LMA and the CN) PBA New Packets via the NMAG PMAG sends buffered packets after receiving SPR message and forwards packets heading to the MN Figure 5 4 Proposed PMIPv6 handover procedure The AAA server returns MNs profile by AAA reply message. This policy profile contains MNs home network prefix. The NMAG informs NAP of MNs home network prefix information by Router Advertisement message

PAGE 69

69 Based on this information the NMAG notifies the PMAG that the MN are moving from the PMAG and asks to establish a tunnel from the PMAG to the NMAG using Neighbour Discovery (ND) procedure defined in IPv6 protocol [ 25 ]. In this ND procedure enhanced neighbour solicitation message (ENS) and enhanced neighbour advertisement (ENA) message are used. The ENS message can be made by adding 2 bits H flag at reserved field and adding the MNs home network prefix and MN -Identifi er at option field in neighbor solicitation message frame. Explanation of the 2-bits H flag is described in T able 5 -2 Table 5 2 H flag of ENS message H flag Description 00 MNs home network prefix and Identifier are not included in the ENS message 01 Only MNs Identifier is included in the ENS message 10 Only MNs home network prefix is included in this ENS message 11 MNs home network prefix and Identifier are included in ENS message By sending the ENS message to the PMAG, the NMAG informs the PMAG of home network prefix and MN -Identifier of MN which has moved to the NMAG and requests the PMAG to perform the L3 handover procedure. If the H -flag in ENS message is 11 the PMAG starts to store packets heading to the MN in its buffer during handover procedure and prepares to establish a tunnel between the PMAG and the NMAG. If the PMAG cannot obtain both of the MNs home network prefix and identifier from ENS message, these procedures for efficient and reliable handover are not performed. The PM AG sends ENA message as an acknowledgment to the NMAG Similarly with ENS, ENA message contains 2bit A flag at reserved field and it can also include MNs home network prefix and MN -Identifier at option field. Frame format of

PAGE 70

70 ENA is same with that of the neighbour advertisement message except A flag. Explanation of the 2-bits A flag is described in T able 5 -3 With ENA with 11 A -flag from the PMAG, the NMAG knows the PMAG starts to do the L3 handover procedure and establishes a tunnel between the PMAG and NMAG. Table 5 3 A flag of END message A flag Description 00 Negative acknowledgement of ENS message because of absence of the MNs home network prefix and Identifier 01 Negative acknowledgement of ENS message because of absence of the MNs home network prefix 10 Negative acknowledgement of ENS message because of absence of the MNs Identifier 11 Positive acknowledgement of ENS message After finishing tunnel setup the NMAG sends stored packet request (SPR) message to ask the PMAG to send stored packets heading to the MN in its buffer to the NMAG if the NMAG is ready for receiving packet. After receiving this message the PMAG does not store packets but forwards packets heading to the MN to the NMAG. This SPR message is introduced to support more efficient commun ication between PMAG and NMAG. If the NMAG receives packet from the CN via the PMAG the NMAG sends multicast proxy binding update (M -PBU) message to the LMA and the CN for registering MN s information. T hen both the LMA and the CN returns proxy binding ack nowledgment (PBA) message to NMAG and the LMA sets a bidirectional tunnel between the NMAG and LMA. This multicast proxy binding update (M -PBU) message can make the CN choose efficient route and avoid the triangle problem. After sending M -PBU message the N MAG forwards packets from the PMAG to the MN. After finishing

PAGE 71

71 r egistration of the MN at the LMN and the CN, packets from CN are delivered to the MN via the LMA and the NMAG or via just the NMAG. Proposed PMIPv6 Protocol 2 In this section a new scheme for PMIPv6 protocol over IEEE 802.11 network to reduce signalling over head cost with help of IAAP (Inter Access Point protocol). In original PMIP protocol the MAG should obtain the MN s profile from the policy store and check the binding information of the M N after L2 handover completion though the MN still stays in the same MAG area. This useless signalling caused by just L2 handover is significant in the host based mobility management protocol. Our proposed scheme make the MAG know whether the MN is moved f rom the new MAG or not after the MN is attached to AP. Then no signalling overhead for the obtaining the MN s profile and checking the binding information occurs in the just L2 handover case. The MN is not involved in the any signalling for supporting IP mobility. The same terminology with PMIPv6 is used and MNs localized mobility is considered. We also consider all APs use IAPP defined in IEEE 802.11f [25] We assume the MAG knows all inform ation which the APs receive simultaneously and these APs know the IP address of the MAG. The MAG sends its address periodically to APs that are connected directly with it. These APs have a distinct buffer that is available only for the address of the MAG. In IEEE 802.11 standard, the handover has typically three distinct phases: discovery, authentication and association. After discover and authentication procedure, a MN send reassociation request message to new AP with previous AP s MAC address. We just co nsider signals between APs in this paper. Upon receiving reassociation request message new AP starts to communicate with previous AP with IAPP. The new AP send IAPP MOVE notify packet

PAGE 72

72 to the old AP. Then the old AP transmits an IAPP MOVE response packet to the new AP. IAPP Version Command : 1 Identifier Length DATA Address Legth Reserved MAC address Sequence Number Length of Context Block Context Block Figure 5 5 IAPP MOVE notify packet format As shown in Figure 5 5 Move notify message has 8 Reserved bits [25] We define 1 -bit A flag to ask old AP to inform address of MAG which is connected with the old AP with IAPP MOVE respond packet. If A flag is 1 the old AP is requested to send MAG s address. If A flag is 0 this packet works as a normal MOVE notify mes sage. Upon receiving MOVE notify packet with 1 A-flag from the new AP, the old AP puts current MAG s address in context block of MOVE response packet and return it to the new AP. Then the new AP transmits this MOVE response packet to the MAG and reassociation response message to the MN. The MAG checks the address of MAG connected with old AP in context block. If the address of MAG in MOVE response message is different with that of current MAG the MAG recognizes that the MN moved from other MAG region an d performs L3 handover procedure. If the address of MAG in MOVE response message is same with that of current MAG the MN performed just L2 handover so no L3 procedures are necessary. Figure 5 -5 shows handover procedure of proposed scheme. In this figure P AP and NAP are in the same MAG1 region otherwise NAP2 is in the MAG2 region.

PAGE 73

73 PAPMAG2 MAG1 Policy Store LMA Profile request Profile Reply Proxy Binding Update Proxy Binding Ack DATA Packets from CN RA with Home Network prefix(HNP) CN Reassociation request Reassociation response Layer 2 Handover MN NAP1 NAP2 IAPP MOVE notify w/ A-flag IAPP MOVE response w/MAG1 address Packets Packets Reassociation request IAPP MOVE notify w/ A-flag IAPP MOVE response w/MAG1 address Reassociation response Tunnel RA with Home Network prefix(HNP) Layer 3 Handover Figure 5 6 The proposed PMIPv6 scheme over IEEE 802.11 network. Analytical model We will explain system and user mobility model used in this paper. A common hexagonal cellular network configuration is used in this paper as shown in Figure 5 7 We assume each hexagonal cell is an AP area and every cell has same shape and size. A subnet served by AR/MAG is composed of several layers of cell. The innermost cell is

PAGE 74

74 called layer 0 which is surrounded by a layer1. And the layer 1 is surrounded by layer 2 and so on. If an outermost layer of a subnet is layern-1 this subnet is called n layer subnet. A layer i has 6i cells where i>0 and an n layer subnet has 3n2-3n+1 cells. Also, it is assumed in this analysis that every subnet has same number of layers. Subnet1 with 3n(n+1)+1 number of AP areas Subnet1(MAG/AR area) Subnet2(MAG/AR area) Subnet2 with 3n(n+1)+1 number of AP areasRa Ra Figure 5 7 Network Configuration The Domain is composed of several subnets with same manner of subnet composition. We define AP area radius Ra for hexagonal cell as the distance from centre to the vertex of cell. To calculate mean AP area/subnet residence time the hexagonal AP area is approximated by a circle which has same area [ 35 ]. The radius of approximating circle for AP area is 0a33 R=R 0.91R 2a (5. 2 ) The radius of approximating circle for nlayer subnet can be expressed as below

PAGE 75

75 233 (331) 2naRnnR (5. 3 ) With a difference with other analysis of mobile IP network [ 28 ][ 31][ 32], cell area and MNs velocity is considered for a factor of mean AP area and subnet residence time in this mobility analysis. We assume MNs are uniformly distributed in the network and the average velocity of an MN is same in all AP areas and subnets. MNs move in strait The cell crossing time is just considered as the AP area/subnet residence time for simple analysis. From [ 3 4] the mean AP area residence time (a) and mean subnet residence time (s) can be derived as follows: 0 aR = 2v (5. 4 ) n sR = 2v (5. 5 ) Where Ro and Rn are radius of approximating circle for AP area and n-layer subnet respectively. v is an average velocity of the MN. To obtain the MNs average number of crossing AP areas/subnets we derive the sessions. We assume incoming sessions follow the Poisson process. Let tc be session inter arrival time which is the time interval between two sessions and ta, ts be a random variable of the AP area residence time and subnet residence time respectively. Let fc(t), fa(t) and fs(t) be the probability density function of tc ta and ts respectively. fa(t) and fs(t) are the general probability density function. Because incoming session are Poisson process, fc(t) it is expressed as

PAGE 76

76 ( ) = c c1 E[t]= (5. 6 ) c is session arrival rate. With these parameters we can derive the probabilities as(k ) wherein MN crosses J AP areas and K subnets respectively during a session inter arrival time tc ,as follows [ 36 ]: ( ) = 1 [ 1 ( ) ] = 0 [ 1 ( ) ]2[ ( ) ] 1 > 0 (5.7) ( ) = 1 [ 1 ( ) ] K = 0 [ 1 ( ) ]2[ ( ) ] 1 > 0 ( 5. 8 ) where *()afs and *()sfs is the Laplace transform of fa(t) and fs(t) respectively. For simplicity of analysis, we assumed that the AP residence time ta and the subnet residence time ts a (1/a) and s (1/s) respectively [ 28 ]. The Laplace transform of fa(t) and fs(t) is expressed as *()a a afs s ( 5. 9 ) *()s s sfs s (5. 10 ) From ( 5. 6 )-(5. 10 ), We have J a (J) J0.ca caca (5. 1 1 )

PAGE 77

77 K s (K) K0cs cscs ( 5. 1 2 ) From ( 5.11)-( 5.12), We can easily derived the MNs average number of crossing AP areas (E(Na)) and subnets (E(Ns)) as follows: ( ) = ( ) = = 0 (5. 1 3 ) ( ) = ( ) = = 0 ( 5. 1 4 ) Since an MN which crosses AR/MAG coverage area also crosses AP area simultaneously we can obtain the MN s average number of crossing AP areas when the MN still stays in the same AR/MAG coverage area (E(Nl) as: as las c E(N)=E(N)-E(N)= (5.1 5 ) Performance Analysis The performance of the MIPv6 with FMDD and proposed PMIPv6 scheme1 are evacuated in the terms of the handover latency and packet loss. T he proposed PMIPv6 scheme 2 is evaluated in the terms of signalling cost. For comparison conventional MIP v6 and PMIPv6 protocols are analyzed as well. We do not consider any security scheme in both host and network based MIP protocols. However the procedure of obtaining the MN s profile from the policy store like AAA server is necessary in network based MIP protocol to make the MN feel in the home network. Also we just consider intra domain handover for network based MIP protocol in this analysis. The parameters used in the analysis are shown in T able 5 4

PAGE 78

78 Table 5 4 Parameters Symbol Description TX Y Transmission delay between node X and Y dX Y A verage number of hops between node X and Y Bw B andwidth of the wired link Bwl B andwidth of the wireless link Lw L atency of the wired link L wl L atency of the wireless link C X Y Transmission cost of packet between node X and Y PC X Processing cost in node X Unit transmission cost over wired link Unit transmission cost over wireless link Analysis of Handover Latency and Packet Loss The handover latency is defined as an interval from the time the MN has lost L2 connection with previous AP until the time the MN receives the first packet delivered from CN in this analysis. The handover latency consists of the L2 handover latency (TL2) and L3 handover latency (TL3). Therefore the total handover latency (HL) can be given : HL= E(Na)TL2 + E(Ns)TL3 (5.1 6 ) The L2 handover is composed of 3 procedures as discovery, authentication and reassociation in IEEE 802. 11standard [24]. In discovery procedure two types of scanning are exist: active and passive scanning. In active scanning the MN try to find a candidate AP with sending probe request message to neighbouring APs. In passive mode scanning the MN listens to periodic beacon from neighbouring APs [27 ]. T his latency for finding candidate AP is called probe delay (Tscan) which consumes almost L2 handover latency The average L2 handover latency varies about 40400ms [2 7 ]. If the MN has selected an AP to move the MN sends authentication request message to the AP and receives authentication response message from the AP. This process is called authentication and the latency for this process is called authentication delay (Tauth .).

PAGE 79

79 After authentication the MN sends reassociation request message to the AP and receives reassociation response message and then L2 handover is completed. This procedure is called reassociati on and the duration of this procedure is called rea s sociation delay (Treas). The L2 handover latency TL2 can be expressed as: TL2 = Tscan + Tauth + Treas ( 5. 1 7 ) We assume passive scanning is used and the MN listens beacon message from 3 neighbour APs. If the APs are con figured only on channel 1,6 and 11 and the beacon interval is 10 msec as describe in [ 26 ], the probe delay Tscan is 90 msec. We define the authentication delay and reassociation delay as the round trip time of IEEE 802.11 control packet between the MN and AP, 2TMN AP(sl). T he oneway transmission delay between nodes X and Y for a message of size s via one hope wireless and dX Y -1 hops wired links is expressed as follows: ---() (-1)XY wlXY wXY wl wsl sp TsLdLdp BB ( 5. 1 8 ) where p is average processing delay of each node The size of IEEE 802.11 control packet and IP mobility management packet is sl and sp respectively. For simplicity of the analysis we assume network is symmetric and all the IEEE 802.11 control packets have same size. Also the size of control packets for IP mobility management is same in this analysis Mobile IPv6. The IP handover latency (TL3) of Mobile IPv6 protocol can be composed of Movement detection latency(TMD), address configuration latency (Tadd.) and location update latency (TU) as shown in Figure 5 -8 [20].

PAGE 80

80 Probe Delay Movement Detection Latency New CoA & DAD Latency Location Update Latency for HA Location Update Latency for CN AP Switching Delay L2 Handover Latency (TL2)IP connectivity latency (TIP) Location Update Latency (TU) tMN-ARtAR-MNtMN-HAtMN-CNtHA-MNtCN-MNL2 handover completed Send RS Receive RA Send NS for DAD Send BU to HA Send BU to CN Receive Back Receive packet from CN tDAD Figure 5-8 Timing diagram of MIPv6 In movement detection procedure the MN sends RS message right after L2 handover completion and receives RA message from the AR. Then TMD is same with 2TMN AR(sp). After movement detection the MN obtains a new CoA. If we assume all MNs obtain a CoA by means of stateful address autoconfiguration and DAD procedure is not necessary in the analysis. The address configuration latency, Tadd. can be roundtrip time between the MN and DHCP server, 2TMN DHCP (sp). If we assume al l MNs obtain a CoA by means of stateless address t he address configuration latency, Tadd. can be the delay for DAD procedure TDAD (1s). With Fast movement detection & DAD (FMDD) scheme t he address configuration latency, Tadd can be a just look up delay i n neighbour cache of AR. Then MN registers its location at HA and CN. Therefore the total handover latency of MIP protocol with DHCP server can be expressed as follows: HL(MIP) = E(Na)[TL2]+ E(Ns)[TL3] (5. 1 9 ) where TL2 = Tscan+4TMN AP(sl) and TL3 = 2 TMN AR(sp) + Tadd. + TMN CN(sp) + TCNMN(sd) The sd is size of data packet. Since there are no special scheme to protect packet loss under ongoing handover in MIPv6 protocol packets sent to the MN are lost during the handover latency. The

PAGE 81

81 number of packet loss is proportional with handover latency. So the total number of packet loss can be given as: PL(MIP) = pHL(MIP) (5. 20 ) where p is average packet arrival rate. Proxy Mobile IPv6 Unlike the host based mobility management protocol, PMIP does not need to perform the movement detection procedure. However the procedure for obtaining policy profile of the MN is inevitable in network based MIP protocol. Probe Delay AP Switching Delay L2 Handover Latency (TL2)tAP-MAGtAP-MN tMAG-AP Send AAA Request Policy Store Access Latency Location Update Latency tMAG-AAAtAAA-MAGtLMA-MAGtMAG-LMA Receive MN-ID Receive AAA Reply Send PBU Send RA & Packet from CN Receive PBA Receive RA & packet L2 handover completed L2 handover start MAG MN Figu re 5 9 Timing diagram of PMIPv6 The MAG gets the MN s profile from the policy store like AAA server then it can send RA message with same home prefix to the MN. Therefore the MN believes it is always on the home domain. The delay for obtaining profile is 2 TMAG AAA(sp). After the MAG has the MN s profile the MAG performs location update to the LMA. The MAG sends PBU message to the LMA then the LMA sends PBA message and data packet to the MAG. T he total handover latency of PMIPv6 can be given as follows: HL(PMIP) = E(Na)[TL2] + E(Ns)[TL3] (5. 2 1 )

PAGE 82

82 where TL2 = Tscan + 4TMN AP(sl) and TL3= 2 TMAG AAA(sp) + TMAG LMA(sp) + TLMA MAG(sd) + TMAG MN (sd) Since the PMIP protocol does not support any packet loss protection scheme the number of packet loss is given as: PL(PMIP) = pHL(PMIP) (5. 2 2 ) Proposed PMIPv6 Protocol 1 In Figure 5 10 the timing diagram of our proposed scheme is illustrated. The NMAG obtains the MN s profile by MN -ID from reassociation request message at L2 handover procedure. After receiving the MN s profile the NMAG sends RA to the MN with home prefix in profile and asks PMAG to establish tunnel by sending ENS message simultaneously The PMAG replies with ENA message and starts forwarding packets to the NMAG after receiving SPR message. Then the MN can receive packets from the NMAG via the PMAG. The MN can receive packets from CN not via the PMAG after location update to the LMA. The weight of packet transmission delay ( =1.2) by tunnel is used in the analysis. The total handover latency of our proposed PMIP protocol can be expressed as follows: HL (P PMIP 1 ) = E (Na) [TL2] + E (Ns) [TL3 Treas] (5.23) where TL2 = Tscan + 4TMN AP(sl) and TL3 = TMN NMAG(sl) + 2TNMAG AAA(sp) + 3TNMAG PMAG(sp) + TPMAG NMAG(sd) + TNMAG MN(sd) After receiving the ENS message from the NMAG the PMAG starts to store packets heading to the MN. So packets are not lost from this point. The number of packet loss is given as: PL (PPMIP 1 a) [TL2] + E (Ns) [TLoss Treas]} (5.24)

PAGE 83

83 where TL2 = Tscan + 4TMN AP(sl) and TLoss = TMN NMAG(sl) + 2TNMAG AAA(sp) + TNMAG PMAG(sp) Figure 5 10 Timing diagram of proposed scheme1 Analysis of S ignaling Cost The signalling cost is composed of the signalling cost caused by inter -AP roaming with staying in the same AR/MAG area (CA) and by inter -AR/MAG roaming (Cs). Then the total signalling cost CT can be expressed as: CT = E(Nl)CA + E(Ns)CS (5.25 ) The inter AP roaming can cause the inter AR/MAG area roaming but not always cause it. If the MN crosses the AP area within the same AR/MAG area the IP configuration and location update procedures are not necessary However the MN in

PAGE 84

84 host based MIP and the MAG in network based MIP do not know whether the location update is needed or not although the MN just performs L2 handover. In host based MIP protocol the MN knows whether it is in the same AR coverage area by receiving RA message from the AR after crossing an AP area. T his procedure is called movement detection. We assume the MN sends RS rather than just waits the RA message right after L2 handover. This movement d etection procedure is also performed in just L2 handover case and same in host based MIP protocols except the FMIP. In FMIP protocol the MN communicates with AR before crossing the AP area to get information of the candidate AR. The impact of inter AP handover in host based MIP protocols are well analyzed in [ 37 ]. In network based MIP protocol, PMIP the movement detection of the MN is not necessary within same domain. The MN recognizes it always stays in the same MAG coverage area. However the MAG has to do location update of MN to the LMA if the MN has been moved from the other MAG area. So the MAG has to obtain the MN s profile from policy store by MN -ID. With the MN s home network prefix the MAG checks whether the MN needs location update or not the n sends the RA message to the MN in just L2 handover case. However the MAG cannot know whether the MN move s from the same MAG region or not the MAG perform s L3 handover procedure s in just L2 handover case In this analysis we just consider signalling cost for L3 handover because L2 handover procedures are same in every protocol T he cost by inter AP roaming with staying in the same AR/MAG area, CA is composed of the cost by the procedure until receiving RA after L2 handover.

PAGE 85

85 The packet transmission cost is proportional to the hop distance and packet size. We use packet size ratio (transmitting packet size/sl) instead of real packet size. The transmission cost of IEEE802.11 control packet between the X and Y via a wired link can be expressed as CX Y = X Y where is the unit transmission cost over wired link. We use as the unit transmission cost over wireless link and the is generally larger than The cost by crossing AR/MAG area CS is composed of the binding update cost (CB) and packet delivery cost (CP). Then CS is expressed as: CS = CB+CP (5.26 ) T he control message for IP mobility management such as RA and BU causes binding update cost CB. The packet delivery cost CP is induced by data packet transmission in handover procedure during ongoing session. The costs of forwarded packet (Cforward) and los t packet (Closs) are considered for the packet delivery cost and the CP can be calculated as follows: CP = Cforward + Closs (5.27 ) W here and are weighting factors and the sum of and is 1. Mobile IPv6 Protocol T he movement detection procedure is always performed after L2 handover in both case of L2 and L3 handover. Otherwise the IP configuration, DAD and location update procedures are required only for L3 handover We assume t he MN sends RS and receives RA from the AR in MIP. T he cost by inter AP roaming with staying in the same AR/MAG area, CA is given by : CA(MIP) = 2 pCMN A R +PCAR+PCMN ( 5.28) where p = sp/sl

PAGE 86

86 T he binding update cost CB are given by : CB(MIP) = 2 P (CMN A R +CMN HA + NCNCMN CN) + P CMN A R + 2PCAR + PCHA+NCNPCCN (5.29 ) where is the weighting factor for multicast transmission of NS in DAD procedure. We define the number of CN, NCN is the number of active MNs and PC is process cost of node. 22 CN 0N23R3n+3n+1 m (5.30) where is density of MN in AP area and m is number of AR in the domain in the n layer subnet The is ratio of active MNs to total number of MNs in AR area [ 31]. In MIPv6 there is no forwarding scheme during handover then the packet delivery cost, CP is induced by just packet loss and given as follows: CP(MIP) = d (CCNPAR+ CPAR MN) p HL(MIP) ( 5.31) where d = sd/sl Proxy Mobile IPv6 protocol The MN in network based MIP does not required to participate in any IP mobility management procedure. After L2 handover the MN does not send any signal on the other hand the MAG should obtain the MN s profile from policy store using MN -ID, MAC address in IEEE802.11 standard from L2 attachment and sends RA message to the MN in both case of inter AP roaming and inter MAG roaming. The location update procedure is required for just inter MAG roaming. T he cost by inter AP roaming with staying in the same MAG area, CA is given by : CA(PMIP) = CAPMAG + P (2CM AG PS + 2CM AG LMA + CMAG MN) + 3 PCMAG + PCPS + PCLMA (5.3 2 )

PAGE 87

87 T he binding update cost CB are given by : CB(PMIP) = CAPMAG + P (2CM AG PS + 2CM AG LMA + CMAG MN) + 3 PCMAG + PCPS + PCLMA ( 5.3 3 ) Since the PMIPv6 has no forwarding scheme during handover the packet delivery cost is just induced by packet loss. The packet delivery cost CP is CP(PMIP) = d (CCNLMA+ CLMA PMAG + CP M A G MN) p HL(PMIP) (5.34) Proposed PMIPv6 Protocol 2 In the proposed PMIP we assume all APs use IAPP. Using this IAPP new AP can get an address of the MAG which is connected with the old AP. T he new AP sends this information to the MAG then the MAG can recognize that the MN performs L2 or L3 handover. So the signalling for obtaining the MN s profile and sending RA message does not occur in the just inter AP roaming case. And L3 handover signalling cost is same with the PMIPv6. The signal flow of proposed scheme is shown at Figure 5 6 T he signalling cost by inter AP roaming with staying in the same MAG area, CA is expressed as follows: CA(P-PMIP 2 ) = CAPMAG + P CMAG MN + PCMAG (5.35 ) T he binding update cost CB is given by : CB(P-PMIP 2 ) = CAPMAG + P (2CM AG PS + 2CM AG LMA + CMAG MN) + 3 PCMAG + PCPS + PCLMA (5.3 6 ) The packet delivery cost CP is CP(P -PMIP 2 ) = d (CCNLMA+ CLMA PMAG + CP M A G MN) p HL(PMIP) (5.37 )

PAGE 88

88 Numerical Result and Comparison This section presents numerical result of the analysis shown in previous section. Parameters and default value used in this analysis are referred from [28, 2 9 ]. Table 5-4 presents these values. Figure 5 1 1. N etwork t opology Figure 511 shows network topology used in the analysis. We assume hop distance from every AR/MAG to HA, CN, AAA/DHCP server and LMA is same respectively. We set a= b=1, c =2 d =5,e=7 ,f=g=10 and h=i=15. Also we define proce ss cost as follows : PCHA = 24,PCCN = 4, PCAAA = 12, PCLMA = 12 and PCMAG/AR = 8

PAGE 89

89 Table 5 -4 Parameters for analysis Parameters for analysis Symbol Value W ired link bandwidth Bw 100 Mbps Wireless link bandwidth Bwl 11 Mbps L atency of the wired link Lw 2 msec L atency of the wireless link Lwl 5 msec IEEE 802.11control packet size sl 48 bytes IP mobility Control packet size sp 96 bytes Data packet size sd 200 bytes P rocess and queuing delay of node p 0.5msec Packet arrival rate p 20 packets/sec Unit transmission cost over wired link 1 Unit transmission cost over wireless link 2 Numerical Result of MIPv6 with FMDD These numerical results show the total handover latency of Mobile IPv6 protocols during an inter -c) as function of several parameters Figure 5 -1 2 illustrates the total handover latency according to subnet layer. Total handover latency is decreased abruptly as subnet layer is increased. In the case of 1 subnet layer, AR/MAG has just one AP. Thus inter AP roaming causes L2 and L3 handover simultaneously. As the number of AP area in subnet is increased the number of L3 hand over is reduced. It means the portion of L2 handover in total handover latency becomes bigger as layer of subnet is larger. If the sub net layer is too large as MNs AR/MAG residence time is longer than session duration, the total handover latency is only composed of L2 h andover latency. The Figure 5 1 2 shows the handover latency of Mobile IPv6 with FMDD scheme and DHCP server is less than 5. 4 sec Otherwise the handover latency of conventional MIPv6 protocol is around 27sec at 1 subnet layer because of a long duplicate address detection delay.

PAGE 90

90 Figure 5 1 2 Mobile IPv6 h andover latency comparison 1 Figure 5 1 3 shows the total handover latency as function of AP area radius. As radius of AP area becomes larger the number of handover is decreased in session duration. Consequently total handover latency is decreased as radius of AP area increases. And the differenc e of latency among each protocol is reduced as AP area radius is increased. The Figure 5 13 shows the handover latency of Mobile IPv6 with FMDD scheme and DHCP server is less than 4.5 sec. Otherwise the handover latency of conventional MIPv6 protocol is ar ound 15sec when radius of AP area is 50m. Figure 5 1 4 presents the relationship between total handover latency and MNs average velocity. The total latency is augmented linearly according to ascent of MNs velocity. The gap of handover latency among protocol is larger as augmentation of MNs velocity When t he average velocity of the MN is 5m/s the gap of handover latency between MIPv6 and MIPv6 with FMDD is around 21.2 sec. 1 1.5 2 2.5 3 3.5 4 4.5 5 0 5 10 15 20 25 30 Subnet layerTotal Handover Latency (sec) MIPv6 MIPv6 W/DHCP MIPv6 W/FMDD

PAGE 91

91 Figure 5 1 3 M obile IPv6 handover latency comparison 2 Figure 5 1 4 M obile IPv6 handover latency comparison 3 50 75 100 125 150 175 200 0 2.5 5 7.5 10 12.5 15 AP area radius(m)Total Handover Latency (sec) MIPv6 MIPv6 W/DHCP MIPv6 W/FMDD 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 5 10 15 20 25 30 Average velocity of MN (m/sec)Total Handover Latency (sec) MIPv6 MIPv6 W/DHCP MIPv6 W/FMDD

PAGE 92

92 Numerical Result of P roposed PMIPv6 S cheme 1 Figure 515 shows the total handover latency of proposed PMIPv6 scheme according to subnet layer. For the comparison this figure presents the handover latency of a conventional PMIPv6 and MIPv6 with DHCP server. Our proposed scheme combined L2 and L3 handover procedures however; it affects mostly L3 handover latency. Therefore improvement of proposed scheme becomes less as subnet layer is increased. Our scheme reduces 12.7 % of total handover latency at 1 subnet layer comparing with PMIP protocol. Figure 5 1 5 PMIPv6 h andover latency c omparison 1 Figure 5 1 6 shows the total handover latency of PMIPv6 as function of AP area radius. As radius of AP area becomes larger the number of handover is decreased in session duration. Consequently total handover latency is decreased as radius of AP 1 1.5 2 2.5 3 3.5 4 4.5 5 2.5 3 3.5 4 4.5 5 5.5 Subnet layerTotal Handover Latency (sec) MIPv6 PMIPv6 Proposed scheme

PAGE 93

93 area increases. And the difference of latency among each protocol is reduced as AP area radius is increased. Figure 5 1 7 presents the relationship between total handover latency and MNs av erage velocity. The total latency is augmented linearly according to ascent of MNs velocity. The gap of handover latency among protocol is larger as augmentation of MNs velocity Figure 5 1 6 PMIPv6 handover latency c omparison 2 Figure 5 1 8 shows the number of packet lost according to subnet layer. As shown in the f igure the number of packet lost is reduced as subnet layer is increased. The number of Packet lost is diminished by buffering & forwarding scheme in the proposed algorithm durin g an ongoing session. Proposed scheme decreased the packet lost as 29.56 % in the case of 1 layer subnet and 11.14% in the case of 2 layer subnet compared with PMIP protocol respectively. Not only Packet loss but also disparity 50 75 100 125 150 175 200 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 AP area radius(m)Total Handover Latency (sec) MIPv6 PMIPv6 Proposed scheme

PAGE 94

94 among protocols is decreased as AP area becomes wider as shown in Figure 5 1 9 Figure 5 20 shows the impact of MNs velocity on packet loss. The number of packet lost is proportional to MNs velocity. The gap among protocols also becomes larger as augment of MNs velocity. Figure 5 1 7 PMIPv6 h andover latency c omparison 3 Figure 5 21 illustrates data packet delay from CN to MN according to packet size during ongoing session. In the proposed scheme the MAG sends multicast PBU message to the LMA and the CN simultaneously. This M -PBU message make CN determine the shortest way to the MAG not via the LMA contrast with PMIP protocol. The maximum transmission unit (MTU) of IPv6 in the Internet is 1280 bytes or greater [ 30 ] The data packet size varies from 200 byte to 1400 byte. The proposed scheme reduces data packet delay around 15ms. 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 1 2 3 4 5 6 7 8 9 10 Average velocity of MN (m/sec)Total Handover Latency (sec) MIPv6 PMIPv6 Proposed scheme

PAGE 95

95 Figure 5 -1 8 PMIPv6 p acket l oss c omparison 1 Figure 5 1 9 PMIPv6 p acket l oss c omparison 2 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 0 20 40 60 80 100 120 Subnet layerNumber of lost packet MIPv6 PMIPv6 Proposed scheme 40 60 80 100 120 140 160 180 200 220 0 20 40 60 80 100 120 AP area radius (m)Number of lost packet MIPv6 PMIPv6 Proposed scheme

PAGE 96

96 Figure 5 20 PMIPv6 p acket l oss c omparison 3 Figure 5 2 1 Packet d elay in s ession t ime 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 0 20 40 60 80 100 120 140 160 180 Average velocity of MN (m/sec)Number of lost packet MIPv6 PMIPv6 Proposed scheme 200 400 600 800 1000 1200 1400 44 46 48 50 52 54 56 58 60 62 64 Data packet sizePacket delay (msec) PMIPv6 Proposed scheme

PAGE 97

97 Numerical Result of P roposed PMIPv6 scheme 2 Figure 5 2 2 shows total signaling cost as function of subnet layer. At subnet layer 1 inter AP roaming always cause inter MAG roaming. Therefore the handover procedure needs much signalling overhead at 1 layer subnet However as subnet layer increased the signalling cost for handover procedure is decreased. At 3 layer subnet PMIP and MIP protocols have almost same signal ling cost and from 4 layer subnet PMIPv6 protocol needs more signalling for handover comparing with MIPv6. At 2 layer subnet proposed scheme reduce 31. 6 % of signalling cost comparing with the conventional PMIPv6. Figure 5 2 2 PMIPv6 c ost c omparison 1 Figure 523 presents the relationship between total cost and MNs average velocity. The total cost is augmented linearly according to ascent of MNs velocity. The gap of total cost among protocol is larger as augmentation of MNs velocity 1 1.5 2 2.5 3 3.5 4 4.5 5 0 2000 4000 6000 8000 10000 12000 Subnet layerTotal cost MIPv6 PMIPv6 Proposed scheme 2

PAGE 98

98 Figure 5 2 3 PMIPv6 c ost c omparison 2 Figure 5 2 4 PMIPv6 c ost c omparison 3 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 Average velocity of MN (m/sec)Total cost MIPv6 PMIPv6 Proposed scheme 2 20 40 60 80 100 120 140 160 180 200 0 2000 4000 6000 8000 10000 12000 AP area radius(m)Total cost MIPv6 PMIPv6 Proposed scheme 2

PAGE 99

99 Figure 5 2 4 shows the total cost as function of AP area radius. As radius of AP area becomes larger the signalling cost is decreased in session duration. Consequently total signalling cost is decreased as radius of AP area increases. And the difference of signalling cost among each protocol is reduced as AP area radius is increased. Conclusion In this chapter, a fast movement detection & DAD (FMDD) scheme for Mobile IPv6 and 2 efficient Proxy Mobile IPv6 protocol s on IEEE 802.11 standard are proposed. The FMDD scheme for Mobile IPv6 use neighbor cache at access router for duplicate address detection procedure which takes at least 1 sec in handover procedure. This proposed FMDD can reduce the latency taken by the movement detection, address configuration and confirmation from the whole handover latency Our propose PMIPv6 scheme1 performs L2 and L3 handover simultaneously to diminish total handover latency and packet forwarding scheme thr ough the tunnel between the NMAG and PMAG by enhanced ND messages is introduced to reduce packet loss during handover. Moreover multicast PBU message make our scheme can decrease data packet delay from the CN to MN. The analysis result shows our scheme1 can reduce the total handover latency and number of packet lost compared to the existing mobile management protocols. The proposed scheme 2 uses MOVE notify and response message in IAPP defined in IEEE 802.11f. Using these messages the MAG can recognize whether the MN moves from different MAG or not. Therefore we can reduce unnecessary signalling cost caused by just L2 handover. The analysis result shows our scheme2 can reduce signalling cost significantly

PAGE 100

100 CHAPTER 6 CONCLUSIONS This dissertation introd uced our proposed schemes and shows the analysis and performance enhancement of home network and mobile network technologies. We provide a comprehensive performance analysis of recent wireless and powerline technologies for home. The recently developed 200Mpbs P owerline modems (HomePlug AV, UPA -DHS and HDPLC) ; IEEE 802.11 g and 802.11n wireless technologies are compared and evaluated. T he results of comprehensive experimental measurements and testing of these technologies with the UDP and TCP protocol s as well as with High Definition TV streaming media at 25 Mbps are shown. We have proposed an enhanced Limited Round Robin with a priority policy for data exchange within a Bluetooth piconet The proposed algorithm utilizes the status of the master and slave t o schedule the slots effectively, yet additional information on the slave is not necessary Also, we introduce EVG ATOR a protocol visualization tool EVGATOR is flexible and configurable to display key trace data as space time diagrams with color coding f or local protocol states This visualization tool can significantly reduce the protocol design process and the time spent on debugging the simulation results We have introduced fast movement detection & DAD (FMDD) scheme for Mobile IPv6 and two efficient network based mobility management schemes over wireless local area network for reducing handover latency and signaling overhead. The F MDD scheme uses neighbor cache for DAD procedure and can reduce handover latency of Mobile IPv6 significantly. To reduce service disruption time and packet loss of Proxy Mobile IPv6 s imultaneous L2 and L3 handover scheme is proposed and modified message in Neighbour Detection procedure is used for handover management. Also,

PAGE 101

101 enhanced binding update message can reduc e the packet delivery delay. To reduce signalling overhead of PMIPv6 we use MOVE -notify message defined in IEEE802.11f. The proposed scheme make s the MAG know whether the MN is moved from the new MAG or not after the MN is attached to AP. Then no signalli ng overhead for the obtaining the MN s profile and checking the binding information occurs in L2 handover case. For performance analysis we developed network configuration and mobility model b y using several parameters to analyze handover latency, packet loss signalling overhead cost. O ur analyzed results show that our proposed schemes have a better performance than conventional mobility management protocols.

PAGE 102

102 LIST OF REFERENCES [ 1 ] IEEE 802.11b Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specification: High-speed Physical Layer Extension in the 2.4GHz Band, IEEE Standard, Sept.1999 [ 2 ] IEEE 802.11g, Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications: Further Higher -Speed Physical Layer Extension in the 2.4 GHz Band, IEEE Standard, June 2003 [ 3 ] Yang Xiao, ,IEEE 802.11N: Enhancements For Higher Throughput In Wireless LANs IEEE Wireless Communications, Volume 12, Issue 6 December 2005,pp. 8291. [ 4 ] Bluetooth Special Interest group, Specification of the Bluetooth System, 1.1b, vol.1 and 2, Feb. 2001 [ 5 ] MOCA, Multimedia over coax alliance, Aug.2007, < http://www.mocalliance.org > [ 6 ] HomePNA, Aug. 2007 ,< http://www.homepna.org > [ 7 ] S. Gavette, HomePlug AV Technology Overview, April 2006 < http://download.microsoft.com/download/a/f/7/ af7777e5 -7dcd48008a0a b18336565f5b/HomePlugAVWP_ShermanGavette.doc > [ 8 ] Digital Home specification white paper ,Universal Power Line Association, May,2006 [ 9 ] High Definition Power Line communication (HD -PLC) http://www.hd plc.org/english/portals/0/20051105_HDPLCMTJ_V4_f(190mbps).pdf [ 10 ] IEEE 802.11a, Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications: High-speed Physical Layer in the 5 GHz Band, IEEE Standard, Sept. 1999. [ 11 ] D. B. Johnson, C. E. Perkins and J. Arkko, Mobility Support in IPv6, IETF RFC 3775, 2004. [ 12 ] E. Rajeev Koodli, Fast Handovers for Mobile IPv6, IETF RFC 4068, 2005. [ 13 ] H.Soliman, C. Castelluccia, K. E. Malki, and L. Bellier, Hierarchical Mobile IPv6 Mobility Management (HMIPv6), IETF RFC 4140, Aug.2005 [ 14 ] S.Gundavelli, Ed., K. Leung, V. Devarapalli, K. Chowdhury, B. Patil Proxy Mobile IPv6, RFC 5213, Aug 2008

PAGE 103

103 [ 15 ] M.K. Lee, R. Newman, H.A. Latchman, S. Katar, and L. Yonge, HomePlug 1.0 Powerline Communication LANs Protocol Description and Comparative Performance Results International Journal on Communication Systems Volume6 Issue 5 April 2003, pp.447 [ 16 ] GatorByte, http://csc.list.ufl.edu/download/GatorByte [ 17 ] Yu -Ju Lin, Sunguk Lee, Haniph A. Latchman, Byungjoo park EVGATOR An Enhanced Visualization Simulator for Networking Protocol Analysis International Journal of Software Engineering and Its Applications, Vol.3, 2009, pp.19 32 [ 18 ] M.Kalia, D.Bansal and R.Shorey, Data Scheduling and SAR for Bluetooth MAC, in Proc. Of IEEE Vehicular Technology Conference, 2000, pp.716720 [ 19 ] M.Kalia, D.Bansal and R.Shorey, MAC Scheduling and SAR Polices for Bluetooth: A Master Driven TDD Pico -Cellular Wireless Syetem in Proc. Of IEEE International Workshop on Mobile Multimedia Communications,1999, pp.384388 [ 20 ] Yang-Ick Joo, Jong Soo Oh, An Efficient and QoS aware Scheduling Policy for Bluetooth, in Proc. Of IEEE Vehicular Technology Conference 2002,pp24452448 [ 21 ] S.Grag, M.Kalia and R. Shorey, MAC Scheduling Policies for Power Optimization in Bluetooth: A Master Driven TDD Wireless System in Proc. Of IEEE Vehicular Technology Conference, 2000, pp.196200 [ 22 ] A.Capone, M.Gerla and R.Kapoor, Efficient polling scheme for Bluetooth picocell s, in IEEE International Conference on Communication, June 2001,pp.19901994 [ 23 ] Sunguk Lee, Haniph A. Latchman, Byungjoo Park ELR R -Enhanced Limited Round Robin Mechanism using priority Policy over Bluetooth Networks International journal of Advanced Science and Technology (IJAST) Volume.6 2009, pp.6170 [ 24 ] Sangheon Pack, Jaeyoung Choi, Taekyung Kwon and Yanghee Choi, Fast Handoff Support in IEEE 802.11 Wireless Networks IEEE Communications Surveys, Vol. 9, No.1 First Quarter 2007, pp.212 [ 25 ] IEEE802.11F Trial -Use Recommended Practice for Multi -Vendoe Access Point Interoperability via an Inter Access Point protocol Across Distribution System Supporting IEEE 802.11 Operation July 2003 [ 26 ] T. Narten, E. Nordmark, and W. Simpson, Neighbor Discovery for IP Version 6, IETF RFC 2461, Dec. 1998.

PAGE 104

104 [ 27 ] A.Mishra, M. Shin, W.Arbaugh, An empirical analysis of the IEEE 802.11 MAC layer handoff process ACM SIGCOMM Computer Communication Review Volume 33 Issue 2 April 2003 pp.93 -102 [ 28 ] C. Makaya and S. Pierre, An analytical framework for performance evaluation of IPv6 -based mobility management protocols, IEEE Trans. Wireless Communication, Volume 7, Issue 3 March 2008 pp :972 983 [ 29 ] J. McNair, I. F. Akyildiz, and M. D. Bender, Handoffs for real -time traffic in mobile IP version 6 networks, in Proc. IEEE GLOBECOM vol. 6, Nov. 2001, pp. 34633467 [ 30 ] S. Deering and R. Hinden, Internet Protocol, version 6 (IPv6) specification, RFC 2460, Dec. 1998 [ 31 ] S. Pack and Y. Choi, A study on performance of hierarchical mobile IPv6 in IP based cellular networks, IEICE Trans. Commun. vol. E87B, no. 3, March 2004,pp.462469 [ 32 ] Shou-Chih Lo; Guanling Lee; Wen-Tsuen Chen; Jen -Chi Liu Architecture for Mobility and QoS Support in All IP Wireless Networks IEEE journal of selected areas in communications, Vol.22, No.4 ,May 2003, pp. 691 705 [ 33 ] Sunguk Lee, Yu-Ju Lin, Haniph A. Latchman, Comparative Performance Analysis of Recent Powerline and Wireless Technologies for Multimedia Home Networking, in Proc the 12th World Multi-Conference on Systemics, Cybernetics and Informatics, Volume III, June, 2008, pp.113118 [ 34 ] Yeung, K.L.; Nanda, S Channel Management in Microcell/Macrocell Cellular Radio System IEEE Transaction on Vehicular Technology, Vol 45, No.4, November 1996, pp. 601 612 [ 35 ] D.Hong and S. S. Rappaport, Traffic model and performance analysis for cellular mobile radio telephone systems with prioritized and nonprioritized handoff procedures, I EEE Trans Veh. Tcchnol., vol. VT -35, no. 3, Aug. 1986 ,pp.77 -91 [ 36 ] Yi -Bing Lin, Reducing Location Update Cost in a PCS Network IEEE/ACM Transcation on Networking, Vol.5, No.1, Feb.1997,pp.25-33 [ 37 ] S.Hwang, Y.Han,S.Min Performance Analysis of IPv6 Mobility S upport protocol over IEEE 802.11 Network IEICE Trans.Commun. vol E87 -B.No.9 Sept.2004 ,pp,2613 2625

PAGE 105

105 [ 38 ] JinHyoeck Choi, DongYun Shin, Router Advertisement Cashing in Access Point (AP) for Fast Router Discovery, draft jinchoi -mobileip -frd -00.txt, June 2002. [ 39 ] Narten, T., Nordmark, E. and W. Simpson, Neighbor Discovery for IP version 6 (IPv6), RFC 2461, December 1998 [ 40 ] S. Thomson, T. Narten, IPv6 Stateless Address Auto -configuration, RFC 2462, Dec. 1998 [ 41 ] V. Srinivasan, G. Varghese, Fast Address Lookups Using Controlled Prefix Expansion, ACM Transactions on Computer System, Vol.17, Feb.1999. [ 42 ] R. Kawabe, S. Ata, M. Murata On Performance Prediction of Address Lookup Algorithms of IP Routers through Simulation and Analysis Techniques, IEEE International Conference on Communications 2002 (ICC 2002)

PAGE 106

106 BIOGRAPHICAL SKETCH Sunguk Lee received hi s Bachelor of Science degree from Electrical Engineering at Kyungpook National University, South Korea in 1998 and his M aster of Science in Electrical and C omputer E ngineering at the University of Florida in 200 4 He received his Ph.D. degree from Electrical and C omputer Engineering at the University of Florida in the spring of 2010. His research interests are in the area of mobility management in IP based mobile net work, w ireless network and p ower l ine c ommunications.