4세대 이기종 네트워크 환경을 위한 에이전트 기반 이동성 및 QoS 관리 방안
Agent-based Mobility and QoS management Schemes for 4G heterogeneous networks
- 주제(키워드) Mobility , QoS , TCP flow control , vertical handoff , 4G
- 발행기관 아주대학교
- 지도교수 임재성
- 발행년도 2006
- 학위수여년월 2006. 2
- 학위명 박사
- 학과 및 전공 정보통신전문대학원 정보통신공학과
- 본문언어 영어
초록/요약
The 4th generation wireless networks will be composed of various kinds of access systems and should provide higher data transmission rates and more diverse multimedia services. Despite the advent of several researches to accomplish these requirements, both guaranteeing reliable quality of service (QoS) and supporting mobility are still difficult. In this dissertation, we define a network architecture based on a local agent called QMA (QoS and Mobility Agent). The QMA supports the mobility and guarantees the QoS in the 4th generation homogeneous and heterogeneous networks. In this 4G-network architecture, we propose a QMA-based mobility management scheme and TCP QoS guaranteeing schemes which are two end-to-end TCP flow schemes, an FS-TCP (Fast Start TCP) scheme, and a DSDA (Delay-based Scheduling with Destination-based TCP flow Aggregation) scheme. First, the QMA-based mobility management provides seamless mobility and reduces packet loss. In order to support both vertical and horizontal handoff, the QMA has four roles; a local registrar, an advertiser, a bi-caster, and a router. The proposed scheme uses SIP (Session Initiation Protocol) signaling messages to establish and maintain sessions. The QMA carries its duty as a home registrar when a Mobile Node (MN) moves between intra-domain subnets. During horizontal handoff, the MN registers its current location information to the QMA instead of an Home Agent (HA). This operation reduces handoff delay in accordance with a long distance registration procedure. Furthermore, the proposed mobility scheme minimizes unnecessary data transmission delay because it starts to send data packets after establishing a communication session. An MN delivers its new location information to a home registrar and a Correspondent Node (CN) via SIP signaling messages when performing vertical handoff. Next, two kinds of TCP flow control schemes are proposed for vertical handoff environments. If an MN conducts vertical handoff between two different access networks such as cdma2000 and WLAN, the retransmission timeout could happen. When an MN moves from WLAN network to cdma2000 network, TCP retransmission timeout will occur in spite of a non-congestion situation because Round Trip Time (RTT) of WLAN is shorter than that of cdma2000. Thus, TCP congestion window size decreases so that TCP throughput degrades. To avoid this undesirable case, we propose two TCP flow control schemes: TCP flow control with a probe packet and TCP flow control with RTT calculation. The first TCP flow control scheme uses probe packet for estimating link delay of the new network after performing handoff. The other method uses bandwidth ratio of each network to update RTT. We compare the proposed schemes with the traditional protocols such as Mobile IP and SIP in terms of handoff delay time, packet loss, location update cost, packet delivery cost, QoS cost, and total cost. Finally, FS-TCP scheme and DSDA scheme are proposed. Both mechanisms are implemented at the transport layer of the QMA agent. The QMA-based FS-TCP scheme is proposed to improve TCP performance in wireless networks. It uses a Snoop inherently based on a local retransmission function which prevents unnecessary congestion control at a sender by hiding wireless packet loss. It also increases the congestion window size of the sender quickly up to the sender…s maximum window size and it works when the amount of buffered packets decreases below the I-threshold. This operation of FS-TCP mechanism is simple but it provides a fast recovery from slow start caused by Snoop retransmission. It is also very efficient in improving the TCP throughput. The guarantee of TCP fairness between users is an important requirement for the users of the network. If scheduling algorithms in a wireless network do not guarantee the level of fairness, the users may suffer low throughput. This dissertation proposes a DSDA scheme implemented at an agent. The agent aggregates TCP flows according to their destination addresses and deals with the aggregated TCP flows in the same way. The agent also performs a packet scheduling according to the transmission delay time of each user. The use of an optimum weighting value according to the delay time improves the fairness. The proposed TCP agent-based scheduling policy can be applicable to several kinds of 4G wireless access networks such as QMAs in the proposed system architecture and wireless access points in mesh networks.
more목차
Table of Contents
CHAPTER 1 INTRODUCTION 1
1.1 Preliminary 1
1.2 Motivation 2
1.3 Dissertation Contour 4
CHAPTER 2 BACKGROUND AND RELATED WORKS 6
2.1 Issues of Mobility and QoS in 4G 6
2.2 Mobility Management? 8
2.2.1 Mobility Issues in Heterogeneous Networks 8
2.2.2 Mobility Management Protocols? 9
2.3 Quality of Service 14
2.3.1 QoS Issues in Heterogeneous Networks 14
2.3.2 QoS Protocols 17
2.3.3 TCP QoS Issues 19
2.3.3.1 End-to-End TCP Flow Control 19
2.3.3.2 TCP Enhancement Mechanism 19
2.4 The Necessity of Agent-based Network Architecture 25
2.5 Summary 27
CHAPTER 3 AGENT-BASED MOBILITY MANAGEMENT SCHEME 28
3.1 System Model and Topology 28
3.2 QMA-based Mobility Management 32
3.3 Performance Analysis 38
3.3.1 Model Description 38
3.3.2 Location Update Cost 41
3.3.3 Packet Delivery Cost 45
3.3.4 Handoff Delay Time and Packet Losses 47
3.3.5 Total Cost for a Session 49
3.4 Performance Evaluation 51
4.5 Summary 60
CHAPTER 4 END-TO-END QOS MANAGEMENT SCHEMES 61
4.1 System Model and Motivation 61
4.2 TCP Flow Control-based End-to-End QoS Management 62
4.3. QoS Cost 68
4.4 Performance Evaluation 70
4.5 Summary 74
CHAPTER 5 AGENT-BASED TCP QOS MANAGEMENT SCHEMES 76
5.1 System Model and Motivation 76
5.2 FS-TCP Scheme 78
5.2.1 Behavior of FS-TCP 78
5.2.2 Performance Analysis 80
5.2.2.1 Snoop 83
5.2.2.2 FS-TCP 86
5.2.3 Performance Evaluation 87
5.3 DSDA Scheme 92
5.3.1 Design Objective and Motivation 92
5.3.2 Chakravorty's Scheme 93
5.3.3 Behavior of DSDA 95
5.3.4 Performance Evaluation 97
5.4 Summary 105
CHAPTER 6 CONTRIBUTIONS AND FUTURE WORKS 107
BIBLOGRAPHY 109
