Вы находитесь на странице: 1из 6

Proceedings of 11th International Conference on Computer and Information Technology (ICCIT 2008) 25-27 December, 2008, Khulna, Bangladesh

End-to-End Mobility Management Solutions for TCP: An Analysis


1 1

Peer Azmat Shah, 2Muhammad Yousaf

Department of Computer Science Mohammad Ali Jinnah University Islamabad, Pakistan peer_shah@hotmail.com 2 Center for Advanced Studies in Engineering (CASE), Islamabad Pakistan myousaf@case.edu.pk
intermediate entity is introduced between MN and CN. These approaches (Mobile IP) do not require any changes to be made on the fixed side but require changes in the existing network infrastructure. This is the main problem as service providers are not willing to change the infrastructure due to additional deployment cost, thus end user suffers. Session layer mobility management approaches are quite favorable but they are only for real time data, so cannot be considered as the best technique. Transport layer mobility management techniques are considered strongest despite the fact that they require changes to the existing protocols like TCP. Mobility management scheme operating at any layer should have following characteristics that attempt to keep the Internet infrastructure unchanged by allowing the end hosts to take care of mobility. There should be no change in the network infrastructure. No need to maintain states in the network [4]. Low handover latency should be involved [5]. Direct communication should be done between MN and CN, avoiding the packet redirection [5]. Application should remain unaware of the handover. Authentication should be simple and highly secure and no third parties should be involved. Mobility management scheme should not affect the TCPs reliability and in-order delivery features. In the last several years variety of end-to-end mobility management schemes have been proposed that include variants of TCP for mobility, SCTP [7] based, HIP based mobility management schemes and Session Layer mobility Management (SLM). SCTP based schemes does not use the existing TCP but they are based on SCTP protocol. Host identity protocol (HIP) [8] is a new secure mobility management protocol that handles mobility by separating the dual role [9] of IP address as end point identifier and locator. In this survey paper we have discussed various end-to-end mobility management schemes for TCP and compared them under common evaluation parameters. We have done an analysis of some selected approaches on the basis of handover delay. Rest of the paper is organized as follows: In section II related works that has been done so far in this domain are discussed. In section III various evaluation criterions have been identified for mobility schemes. Section IV covers a detailed survey of existing solutions for end-toend mobility management. In section V a comparison is made among different techniques and in section VI conclusions is made.

Abstract - Todays mobile devices are equipped with multiple network interfaces belonging to different access networks, these interfaces may be used in different requirements. A network connection should be established using the best possible interface. During communication, characteristics of a network may change or a new access network is reached by Mobile Node thus an already established connection should be moved from one interface to another or parallel transmission is done over two or more connections. In the last several years variety of mobility management solutions have been proposed that operate at different layers of the communication protocol stack (e.g link layer, network layer, transport layer and application layer), but the network and transport layer mobility management solutions are more mature. In this paper we have discussed various end-to-end mobility management solutions for TCP. We have also identified a number of evaluation parameters and on the basis of these parameters different mobility management techniques are analyzed. Index Terms - Mobility Management, vertical handover, transport layer, TCP, end-to-end

I. INTRODUCTION Next generation mobile communication system (4G) will provide all time access and seamless mobility across different heterogeneous networks like WLAN, WiMAX, UMTS and WWAN etc. In wireless networks, mobility of a Mobile Node (MN) can be defined as changing the point of attachment without loosing its ability to communicate in the network with correspondent node (CN). Whenever a MN changes its point of attachment, handover occurs. Handover is divided into two types: horizontal and vertical. A horizontal handover is made between different base stations within the same link layer technology; it is usually due to the geographical movement of a MN in a network. A vertical handover means that MN hand-off from one underlying access network to another. A common goal [1] among all the approaches for mobility management is that mobility should not break the existing connections between two hosts and also the MN should not become unreachable for future connections. Mobility can be handled at different layers [1] of the traditional TCP/IP protocol stack but in any case link layer support is necessary for providing meaningful information to initiate the handover. It can be handled at Network layer (MobileIP [2] etc), Transport layer (SCTP, variants of TCP etc.) and at session layer (SIP, SLM etc.). At each of these layers mobility management schemes have strengths and weaknesses. At Network layer mobility management has several shortcomings [3], e.g. an

1-4244-2136-7/08/$20.00 2008 IEEE

696

II. RELATED WORK Mobility management schemes have been discussed and analyzed in the literature at various layers and in varying contexts. For instance in [3] an empirical analysis of handover performance for SIP, Mobile IP, and SCTP Protocols is made but SIP and MobileIP are not based on end-to-end philosophy. In [5] a comprehensive performance analysis of Mobile IP, TCP-Migrate, and SIP is done with the derivation of mathematical model, but the analysis is not made for TCP based connections. [6] has discussed various transport layer mobility management solutions and classified them into different classes on the basis of their functionality. In [10] handover performance is compared for HIP and Mobile IPv6. Different protocols for improving TCP performance during mobility are discussed in [11]. A comparison of host mobility for IP networks using TCPMigrate, HIP and MobileIP is done in [12] but it does not consider end-to-end mobility. All these works consider some mobility management solutions that are based on different protocols. In this paper we have discussed solutions for endto-end mobility management for TCP based communication. A comprehensive analysis is made for each solution under some common parameters and strengths and weaknesses are identified on the basis of these parameters. III. EVALUATION PARAMETERS When a MN moves from one location to another then the mobility management scheme should be able to manage communication without disruption. It should handle handover and thereafter sequenced delivery of data. We have selected a number of evaluation parameters and divided them into three classes. - Handover related issues - Infrastructure and Design of TCP/IP model related issues - Performance related issues A. Handover related issues In this class those parameters are included that are related to the type of handover and handover cost etc. Soft handover: A soft handover allows the MN to continue communication without any disruption. Thus soft handover can be defined as make-before-break. Handover overhead: During the handover some control messages have to be exchanged between the two communicating nodes. These messages are transmitted and processed by the communicating parties. The number of these control messages varies from mobility scheme to scheme, and may create overhead. Low layers awareness: A mobility scheme that takes information about signal strengths etc. from the lower layers, for handover decision, is said to be Low layer aware. Usually this information is provided by the link layer through Media Independent Handover (MIH) triggers (802.21) [30]. Willful handover: User centric approach [13] can be said as the willful handover. The handover must be according to the requirements of the user. It is possible that there may be an

overlapping area of the two or more networks i.e. strong signals of different access technologies may be received by the MN, but user may not want to do handover. Handover latency: It is the time elapsed after the MN receives the last packet in the old network until it receives the next packet in the new network. B. Infrastructure and Design of TCP/IP model related issues In this class, parameters regarding the network infrastructure and the current implementation of TCP/IP model are included. Changes in the current implementation of TCP: Some mobility management schemes require the changes in the current implementation of TCP means that they do not use the conventional TCP control and data signaling. These schemes may be termed as the TCP-variants for mobility management. Requires changes at both ends: A mobility scheme may require changes in the communication protocol stack at either end or at the both ends. Support from Network: A mobility management scheme may require support from the network for control and data signaling, e.g. use of gateways, proxies, mobility agent etc. Cross layer: Cross layer optimization may be used by some approaches to handle mobility. C. Performance related issues Here the parameters that regard the extra functionality and performance of handover are discussed. Location updates: Whenever a MN moves into a foreign network and gets a new IP address, then the new nodes wanting to communicate with this MN would not be able to locate it. For this purpose location updates are required. Multi-homing: Todays mobile devices come up with multiple network interfaces. Thus it is possible for a mobile device to have multiple IP addresses. These IP addresses may be obtained from different subnets. Using multi-homing [14] an association between MN and CN can be spanned across multiple IP addresses. Bandwidth Aggregation: When a multi-homed MN has access to the coverage areas of different wireless networks, it can use multiple interfaces simultaneously for sending and receiving data [14]. Some mobility management schemes use multiple TCP connections between MN and the CN, in this situation a scheduler is required for efficient bandwidth aggregation on parallel connections. Simultaneous movement: It may be possible that two communicating nodes may move simultaneously. A mobility management scheme is said to be simultaneous movement aware if it supports simultaneous movement of MN and CN. Application transparency: A mobility management scheme is application transparent when applications are unaware of the handover process. Security issues: Different mobility management schemes

697

have security flaws and/or in some schemes existing security solutions are implemented which do not meet the requirements of that scheme. IV. MOBILITY MANAGEMENT SOLUTIONS FOR TCP In this section we have discussed various mobility management solutions for TCP. Strengths and weaknesses of each scheme are highlighted. Table.1 shows the comparison among different schemes. A. Freeze TCP In Freeze TCP [15] an ongoing TCP connection is freezed or stopped when the signal strength is fading by advertising a zero window size to CN, thus stopping all transmission before handover. After handover, when MN has moved to some other location and reconnected, the existing TCP freezed connection is unfreezed by sending a message. As soon as the connection is re-established, MN sends Triplicate ACKs for the last data segment received before handover. This scheme reduces the chances of packet loss during handover at the cost of higher delays. It does not require any intermediaries like proxy, gateway etc. No changes are required in TCP protocol at CN but needed at MN. It does not support multi-homing but requires support from lower layers for handover. No option for bandwidth aggregation as there are no options for multi-homing and parallel connections. B. TCP-Migrate In TCP-Migrate [16] [17] for connection migration author has proposed a Migrate option for TCP that allows an existing TCP connection to be migrated by either host from an old IP address to a new IP address. This TCP connection migration is done by exchanging newly introduced TCP SYN with Migrate option. In migrate option a token (identifier) is included that identifies the previously established connection. This token is negotiated during initial connection establishment using migrate-permitted option. This scheme is a complete end-to-end mobility management solution but the handover involved is hard handover. Just like Freeze-TCP the connection is stopped. It does not require any changes in the network infrastructure but TCP at both ends needs to be changed. No option for multi-homing and bandwidth aggregation. TCP-Migrate handles encrypted traffic and can be implemented with the existing security solutions. This scheme lacks simultaneous movement support and does not preserve location privacy. An enhanced version of TCP-Migrate with simultaneous movement support is proposed in [18]. C. TCR-Redirection For continuous operation on Mobile Node, TCP-R [19] revises the pair of IP addresses and port numbers in the existing connection, whenever there is a change in the MNs IP address. For new connection establishment TCP-R performs two operations: 1.check whether CN is TCP-R aware

2.exchange authentication keys After getting a new IP address, MN sends a redirect message to CN. On receiving the redirect message CN sends authentication request to MN. MN replies with connectionauthentication message. CN checks the connection authenticator, if found correct then revises the pair of addresses of existing TCP connection. In the meanwhile MN also revises its pair of addresses. The scheme can be implemented as a complete mobility solution and is application transparent. It does not require any support from lower layers for connection migration. DNS is proposed for location management. TCP-R does not support multi-homing and bandwidth aggregation. D. VA-TCP Vertical handoff Aware-TCP (VA-TCP) [20] is an end-toend mobility management scheme in which a new abstraction layer is inserted between Network layer and Transport layer for connection continuity. This new layer adds some new messages to support vertical handover and also maintains a Local Connection Translation (LCT) table for maintaining a relationship between original connection information and the new connection. When a MN roams in a new network, then VA-TCP dynamically estimates the connection parameters from the information provided by the lower layers, so that it can fit into the new environment. VA-TCP is designed to achieve TCPs best performance for end-to-end mobility management that requires changes in the protocol stack by introducing an abstraction layer. Low layer triggers are required for handover initiation. No option for multi-homing and thus no aggregate bandwidth management is required. Handover is not done according to user requirements but is done according to the new network conditions. E. MSOCKS Maltz et.al proposed MSOCKS [21], Transport layer mobility management architecture based on TCP Splice. MSOCKS is built around a proxy inserted in the path of MNCN. Connection from MN to CN is split at proxy and two interfaces are made at proxy. Each interface has its own IP address. Scheme introduces a RECONNECT message that is sent to the Proxy upon getting a new IP address in a new network. MSOCKS enabled MN creates a new socket and sends a RECONNECT message to Proxy with connection ID of old session. Proxy establishes new connection and sends a reply message to MN. Upon receiving reply MN resumes transmission. Mobility in MSOCKS is limited to the coverage area of Proxy as location management is done by Proxy. It also needs low layer information for handover initiation. A node may have multiple IP addresses but communication is done only through one interface so no option for bandwidth aggregation. MSOCKS requires changes in the current implementation of TCP but only at the MN and Proxy. This protocol cannot work with IP layer cryptographic techniques like IP-SEC.

698

F. Tsukamotos scheme with Multiple TCP Connections Tsukamoto proposed a new mobility management scheme that modifies the transport protocol for handling multiple connections [22] during handover process. A cross layer approach is used that detects any changes in the status of wireless networks and selects the best optimal interface. A Handover Manager (HM) is implemented at transport layer that detects the changes in the status of wireless networks. Link layer information is sent to the HM at transport layer bypassing the intermediate layers. HM using this information can detect a link-up or link-down state of a wireless network. If during communication HM detects a link-up state for a new wireless network on a new interface, then it starts a new TCP connection via that interface and starts parallel transmission using two TCP connections. For each connection end-to-end performance measures are calculated and compared. On the basis of this comparison handover takes place. In this scheme no details of the control and data flow are given. It is not mature for implementation and a lot of work needs to be done. No provision of willful handover is there. In this scheme bandwidth aggregation can be possible but author has not given any details about this. G. pTCP pTCP [24] [25] is a transport layer solution for multi-homed MN. It creates and maintains one TCP state for each network interface. TCP-virtual (TCP-v) pipes are created for each active interface in a pTCP connection. Whenever a new pTCP socket is opened by application for a connection then a TCP-v pipe for that connection using active interface is created. Application in the MN opens a pTCP socket with CN and pTCP creates the first TCP-v pipe. This pTCP with only one pipe behaves as the same as TCP. When MN moves to a new network and gets a new IP address then pTCP creates a new pipe but does not closes the previously

established pipe. pTCP provides the reliable and sequenced delivery just like the simple TCP. Both ends need to be pTCP aware. No support is required from the network like proxy etc. It does not support willful handovers but bandwidth aggregation is performed [26] for effective throughput. H. I-TCP Indirect TCP (I-TCP) [23] proposes to split the TCP connection at the mobility support station that is typically a gateway/MSR. Two TCP connections are used, one is an ITCP connection between MN and gateway and the other is regular TCP between gateway and CN. When a MN moves and gets a new IP address then only the TCP connection between MN and gateway is changed and replaced by the new TCP connection. Hard handover is done and protocol requires changes at the MN. No option for multi-homing is available. Network support is required with gateway/MSR. I. SLM Session Layer Mobility management (SLM) [28] is a framework that supports mobility management at session layer. It extends the TCP/IP model by adding session layer. This layer operates above TCP and switches data streams between multiple connections. The end-to-end network path is divided into three separate paths. Two paths are between the applications on the two hosts and socket connectors and the third path between the socket connections of two hosts. Location management is done through User Location server (ULS) in the new network. In this scheme both ends need to be SLM aware and support is required from the network in the form of proxy / User Location Server (ULS). Multi-homing is available and bandwidth aggregation can be done. This scheme also supports simultaneous movement of two nodes. No option is available for willful handover.

TABLE. 1 Comparison of different mobility management techniques Y: Yes N: No


Evaluation Parameters Low layers aware Willful handover Requires changes in TCP Changes required on both sides Multi-homing Bandwidth aggregation Requires support from network Cross layer approach Location updates Application transparency Handover Security issues Handover overhead Handover latency Simultaneous movement aware Freeze-TCP Y N Y N N N N N N Y Hard N N Y N TCPMigrate Y N Y Y N N N N Y Y Hard N N Y N TCP-R N N Y Y N N N N Y Y Hard N N Y Y VA-TCP Y N Y Y Y N N N Y Y Soft N N N Y MSOCKS Y N Y N Y N Y N Y Y Hard Y N Y N Tsukamoto TCP Y N Y Y Y N N Y Y Y Soft N N N N pTCP Y N Y Y Y Y N N Y Y Soft N N N Y I-TCP Y N Y N N N Y N N Y Hard Y N Y N SLM Y N Y Y Y Y Y N Y Y Soft N N N Y HIP based Y N N Y Y N N N N Y Soft N N N Y EMF Y Y N Y Y Y N Y Y N Soft N N N Y

699

J. HIP base mobility management techniques In current Internet IP address is used as both topological locator and an end point identifier. Host Identity Protocol (HIP) [8] separates the dual role of IP address as end point identifier and locator. It introduces a Host Identifier (HI) used as end point name at transport layer and IP address is used as locator at network layer. This is termed as a new layer - Host Identity Layer, which is seen as a 3.5 layer [9]. In this approach multi-homing is allowed but parallel transmission on two links is limited to the fact that when there is data flowing on one link, control packets for handover flow on other link. Therefore no explicit mechanism for bandwidth aggregation has been proposed so far. Willful handover is not considered and both end need to be HIP aware. K. EMF End-to-end Mobility management Framework [29] is a new idea proposed for continuous transmission of data on multiple links between MN and CN. In this scheme cross layer architecture is proposed in which Host Agent gets information from lower layers through MIH triggers [30] and passes to the User Agent on application layer and to Control Agent on transport layer. Host Agent takes handover decision on the basis of information provided by User agent and lower layers. Data handler module is present that is responsible for soft handover and in-order delivery of packets from different TCP connections to the application. In each TCP connection traditional TCP provides reliable data delivery, congestion control and flow control mechanisms. Multiple TCP connections are established under a single EMF association for transmission of single stream of data between MN and CN. In this approach no support is required from the network but changes are to be made on the both sides in transport layer. Willful handover is also possible that makes this approach unique from others. This scheme inherently provides bandwidth aggregation but no details are given. V. ANALYSIS OF DIFFERENT APPROACHES In any mobility scheme handover delay/latency and throughput degradation are the major factors to be determined. Handover latency depends upon number of handover control messages, their number of retransmissions, and RTT. Throughput degradation depends upon handover latency and the time for TCP slow start. TCP slow start time: Assuming no packet loss and TCP doubles its congestion window on each RTT, [5] has defined the time for TCP to reach steady state as:

Handover latency: Assume that three control messages are being transmitted for handover. Message1 (MSG1) with end-to-end packet loss probability p1 is sent (i - 1) 0 times unsuccessfully and i th time successfully. MSG2 with end-to-end packet loss probability p2 is sent (j - 1) 0 times unsuccessfully and j th time successfully and MSG3 with end-to-end packet loss probability p3 is sent (k - 1) 0 times unsuccessfully and k th time successfully and on each retransmission doubling the RTO, handover latency for control messages as explained by [5] is
Latencyhandover = 1.5RTT + 2 a RTO + 2 b RTO + 2 c RTO
a =0 b=0 c =0 i 1 j 1 k 1

Or it may be written as Latency handover = 1.5 RTT + Total handover delay will be:

RTO 2i + 2 j + 2 k 3

(2) (3)

Thandover = Latency handover + TL2 + T

Where TL2 is the time required for layer2 handover, T is the time required for the acquisition of new IP address in the new network. Throughput degradation time: It is the sum of total handover delay and time to reach steady state for TCP. From eq (1) and (3) as defined in [5] (4) To compare the performance of different approaches, we consider the following values for different parameters as defined in [6]: TL2= 10 ms T = 20 ms. TL2 and T are defined in (3). End-to-end packet transportation delay = 100 ms, then RTT = 2x100=200 ms. RTO can be taken as RTO = 1.5xRTT i.e. RTO = 300 ms. We plotted the graph for handover latency by using different number of retransmissions of handover control messages. In Fig.1 a comparison is made for schemes having three control messages (pTCP, TCP-Migrate, HIP) and two control messages (VA-TCP).
Scheme having three control messages Scheme having tw o control messages Handover latency 500 400 300 200 100 0 0 2 4 6 8 10 12 No. of re transm iss ion of control m es s ages

TThput Deg

= Ts + Thandover

Ts = [log 2 (CWs + 1)] RTT Here CWs = Congestion window at steady state Ts = Time to reach steady state

(1)

Fig.1. Handover Latency comparison for schemes having two and three control messages

The results show that greater the number of control messages a scheme has, greater will be the handover latency and vice versa. Similarly in Fig.2 a comparison is made for a single technique (pTCP) by doubling the RTT and RTO values and

700

found that greater values of RTT and RTO leads to longer handover delay.
pTCP w ith RTT=200 ms, RTO=300 ms pTCP w ith RTT=400 ms, RTO=600 ms Handover latency 1000 800 600 400 200 0 0 2 4 6 8 10 12 No. of retransm ission of control m essages

Fig.2. Handover Latency comparison with respect to RTT and RTO

Throughput degradation occurs when there is no provision for parallel TCP connections i.e. for handover initiation connection is stopped or freezed and whenever data is sent on new interface TCP will start from slow start and will take time Ts as defined in (1) to reach at steady state thus resulting in throughput degradation. Whereas in schemes like pTCP, EMF data can be transmitted over multiple interfaces enjoying the capability of bandwidth aggregation, thus there is no throughput degradation. VI. CONCLUSION In this paper we have surveyed various mobility management solutions for TCP and evaluated them under common parameters. Each technique has strengths and weaknesses. A tabular comparison is made highlighting these strengths and weaknesses. Finally an analysis of some approaches is done on the basis of handover delay and throughput degradation and it is found that handover latency depends upon the number of control messages an approach uses for handover initiation, the RTT and RTO values and the link conditions. Similarly throughput degradation occurs when there are no options for bandwidth aggregation. Solutions that can make use of parallel TCP connections between communicating nodes, like pTCP, EMF, Tsukamotos TCP, should have lower handover latency and there should be minimum throughput degradation time for such schemes. In our opinion a solution for mobility management will be comprehensive if it supports multihoming with bandwidth aggregation option and is application transparent. It may take low layer support for handover and should provide willful handover with simultaneous movement of MN and CN having low handover latency and no security issues. Also solution must not depend upon network infrastructure (proxies, gateways, MSR). REFERENCES
[1] Wesley M. Eddy At What Layer Does Mobility Belong? Communications Magazine, IEEE, Oct. 2004, Vol: 42 pp: 155- 159 [2] Charles E. Perkins Mobile Networking Through Mobile IP Internet Computing January/February 1998 (Vol. 2, No. 1) pp. 58-69 [3] Sherali Zeadally, Farhan Siddiqui An Empirical Analysis of Handoff Performance for SIP, Mobile IP, and SCTP Protocols Wireless Personal Communications (2007) 43:589603

Christopher Ross Mobility in the Transport Layer Mobility Without Touching the Network Siemens Mobile Networks magazine [5] Shantidev Mohanty, Ian Akyildiz Performance Analysis of Handoff Techniques Based on Mobile IP, TCP-Migrate, and SIP IEEE Transaction on computing, July 2007 (Vol. 6,) pp. 731-747 [6] Atiquzzaman, M. Reaz, A.S Survey and classification of transport layer mobility management schemes PIMRC 2005. page(s): 21092115 Vol. 4 [7] R. Stewart et. al. RFC 2960 Stream Control Transmission Protocol October 2000 [8] R. Moskowitz RFC 4423 Host Identity Protocol (HIP) Architecture [9] Joseph, Jidong Wang HIP Based Mobility Management for UMTS/WLAN Integrated Networks [10] Jokela, P. Rinta-aho, T. Handover performance with HIP and MIPv6.Wireless Communication Systems, 2004 page(s): 324- 328 [11] H. Elaarag Improving TCP performance over mobile networks ACM Computing Surveys Volume 34 , Issue 3 Pages: 357- 374, 2002 [12] T. R. Henderson Host Mobility for IP Networks: A Comparison Network, IEEE Nov.-Dec. 2003 Volume: 17, Issue: 6 pp: 18- 26 [13] A. Calvagna, G. D. Modica A user-centric analysis of vertical handovers Proceedings of the 2nd ACM international workshop on Wireless mobile applications and services on WLAN hotspots Philadelphia, PA, USA Pages: 137 - 146 [14] K. H. Kim, Y. Zhu A receiver-centric transport protocol for mobile hosts with heterogeneous wireless interfaces Wireless Networks Volume 11, Issue 4 (July 2005) Pages: 363 382 [15] T. Goff Freeze-TCP: A true end-to-end TCP enhancement mechanism for mobile environments INFOCOM 2000. page(s): 1537-1545 vol.3 [16] Alex C. Snoeren and Hari Balakrishnan An End-to-End Approach to Host Mobility 6th ACM/IEEE MobiCom 2000 [17] Snoeren, A.C. Balakrishnan, H. Kaashoek, M.F. Reconsidering Internet mobility This paper appears in: Hot Topics in Operating Systems,2001. May 2001 page(s): 41- 46 [18] Yi Wu; Yanqun Le; Dongmei Zhang An Improved TCP Migrate Scheme with DNS Handover Assistant for End-to-End Mobility ICC apos;07. 24-28 June 2007 Page(s):1923 1928 [19] Funato, D. et al. TCP-R: TCP mobility support for continuous operation Network Protocols, 1997. Page: 229-236 [20] Yu-Chieh Lin VA-TCP: a vertical handoff-aware TCP Symposium on Applied Computing 2007 Seoul, Korea, Pages: 237 - 238 [21] Maltz, D.A. Bhagwat, P. MSOCKS: an architecture for transport layer mobility INFOCOM '98. page(s): 1037-1045 vol.3 [22] Kazuya Tsukamoto et al. Vertical Handover Management Scheme Using Multiple TCP Connections for Heterogeneous Networks WiNTE CH07, September 10, 2007, Montr al, Q ubec, Canada. [23] A. Bakre I-TCP: indirect TCP for mobile hosts Proceedings of the 15th International Conference on Distributed Computing Systems [24] Hung-Yun Hsieh An end-to-end approach for transparent mobility across heterogeneous wireless networks Mobile Networks and Applications Volume 9 , Issue 4 (August 2004) Pages: 363 - 378 [25] H. Y. Hsieh and R. Sivakumar pTCP: An End-to-End Transport Layer Protocol for Striped Connections Proceedings of 10th IEEE International Conference Network protocols Nov. 2002 pages: 24- 33 [26] Hung-Yun H. and R. Sivakumar A transport layer approach for achieving aggregate bandwidths on multi-homed mobile hosts Wireless Networks Volume 11 , 2005 Pages: 99 - 114 [27] H. Sivakumar, S. Bailey, and R. L. Grossman, Psockets: The case for application-level network striping for data intensive applications using high speed wide area networks in Proceedings of Supercomputing 2000. Dallas, TX: November 2000, pp. 240-246. [28] Landfeldt, B.; Larsson, T.; Ismailov, Y.; Seneviratne, A. SLM, a framework for session layer mobility management Computer Communications and Networks, 1999. Proceedings. Eight International Conference Volume , Page(s):452 - 456 [29] M.Yousaf and A. Qayyum On end-to-end- Mobility Management in 4G Heterogeneous Wireless Networks Proceedings of IEEE 2nd International Networking and Communication Conference (INCC2008) pp- 118-123 May 2008. [30] George Lampropoulos et.al Media-Independent Handover for Seamless Service provision in Heterogeneous Networks IEEE Communication Magazine January 2008.

[4]

701

Вам также может понравиться