Elsevier

Computer Communications

Volume 26, Issue 6, 15 April 2003, Pages 542-551
Computer Communications

Performance evaluation of wireless TCP with rerouting in mobile networks

https://doi.org/10.1016/S0140-3664(02)00145-7Get rights and content

Abstract

We study the performance of two different classes of wireless Transmission Control Protocol (TCP) schemes—the split connection TCP and the snoop protocol TCP under four rerouting schemes—full rerouting, partial rerouting, cell forwarding and virtual tree rerouting. The wireless TCP schemes are compared based on the source throughput, average round trip delay and cumulative source disruption time. We analyze the performance of these wireless TCP schemes and their interaction with rerouting.

Introduction

Transmission Control Protocol (TCP) is one of the most widely used transport layer protocols. In a mobile environment, TCP cannot discriminate between the packets dropped due to lossy links, intermittent connectivity, handoffs and rerouting with those dropped due to congestion in the network. TCP reacts to packet losses as it would in a wired environment. It drops its transmission window size before retransmitting packets and initiates congestion control or avoidance mechanism like the slow start algorithm and resetting its retransmission timer. This results in an unnecessary degradation in system performance [7], [8]. In the past, several proposals have been made to alleviate this problem. However, these proposals do not take into account the type of rerouting, viz. full rerouting, partial rerouting, tree based rerouting or cell forwarding [19] after handoff. We believe that the work presented in this paper is the first to study the effect of different kinds of rerouting on the performance of wireless-TCP schemes. Proposed TCP schemes for mobile networks include link-level retransmissions [7], split connection TCP [6], fast retransmit [7], snoop protocol TCP [7], [8] and explicit loss notification [7]. These techniques are discussed in more depth in Section 1.2.

TCP is a protocol used along with the Internet Protocol (IP) to send data in the form of message units between computers over the Internet. While IP takes care of handling the actual delivery of the data, TCP takes care of keeping track of the individual units of data (called segments) that a message is divided into for efficient routing through the Internet. For example, when an HTML file is sent from a Web server, the TCP program layer in that server divides the file into one or more segments, numbers the segments, and then forwards them individually to the IP layer. Although each packet has the same destination IP address, it may get routed differently through the network. At the other end (the client program), TCP re-assembles the individual segments and forwards them to the user as a single file. TCP is a connection-oriented protocol, which means that a connection is established and maintained until such time as the message or messages to be exchanged by the application programs at each end have been exchanged. TCP is responsible for ensuring that a message is divided into the segments and for re-assembling the segments back into the complete message at the other end. In the ISO/OSI network model, TCP is at the transport Layer.

TCP is a means for building an end-to-end reliable byte stream on top of the unreliable packet IP. TCP is the protocol that supports nearly all Internet applications. TCP entities use the sliding window protocol [24] to communicate. The basic method of operation involves

  • wrapping higher level application data into segments

  • wrapping the segments into IP datagrams

  • associating port numbers with particular applications

  • associating a sequence number with every byte in the data stream

  • exchanging special segments to start up and close down a connection between two hosts

  • using acknowledgments and timeouts to ensure reliability of data transmission.

As mentioned earlier, TCP does not adapt well for wireless and mobile environments. Proposed TCP schemes for mobile networks can be classified under the following categories [7], [16]:

  • Link-Level protocols. These protocols hide the packet loss on the wire-less link by performing local retransmission at the link-level and forward error correction over the wireless link. The two main approaches adopted by these protocols include forward error correction (FEC) and retransmission in response to Automatic Repeat Request (ARQ). Examples of link-level protocols tailored for wireless links include the AIRMAIL [4]. If the link level protocol uses the TCP acknowledgments effectively without generating its own acknowledgment, some overhead is minimized. This phenomenon is called as link-level TCP awareness. The snoop protocol [8] is an example of such a scheme. In case of the snoop protocol, a snoop module at the base station caches packets and retransmits if it thinks that data has been lost or corrupted.

  • Split connection. The TCP connection is split at the base station: a regular TCP connection between the non-mobile host and the base station and a modified TCP connection between the mobile host (MH) and the base station. By doing so, data flow on the wireless link is isolated from the flow on the wired connection. Indirect-TCP [6] is an example of the split connection approach.

  • Fast retransmit. The receiver sends a certain number of duplicate acknowledgments informing the sender the packet loss is not due to congestion. The sender then does a fast retransmission of the packet without resorting to congestion control measures.

  • Explicit loss notification. Explicit acknowledgments from the base station are sent to indicate if the loss is due to congestion.

  • End-to-end protocols. These schemes use TCP connections between the source and the MH. However, in order to make the TCP sender handle segment losses, they use either selective acknowledgments (SACKs) to allow the sender to recover from multiple segment losses in a window or an explicit loss notification (ELN) to inform the sender of the segment losses. TCP Reno, TCP Reno with SACK and TCP Reno with SACK and SMART [15] are examples of end-to-end protocols. In the SMART scheme, acknowledgments contain the cumulative acknowledgement and the sequence number of the segment that caused the receiver to generate the acknowledgement.

Mobile networks involve MHs and an underlying wired network consisting of base stations and intermediate routers. Communication is between two entities—the source and the destination, which can either be static hosts or MHs. Communication between the source and the destination involves establishing a path between the base stations in charge of the source and the destination. Base stations are fixed entities that have a region of influence known as a cell. The MHs move from one cell to another and register with the base station of the cell they move into. This process of transferring the control and responsibility for maintaining communication connectivity is called a handoff. In order to maintain communication connectivity, packets are rerouted from the source to the destination base station. The manner in which the route is re-established is commonly known as a rerouting. It should be noted that TCP is a transport layer protocol while the rerouting is a network layer function and hence they are independent of each other.

In this paper, we study the split connection TCP and the snoop protocol TCP under different rerouting schemes—full, partial, cell forwarding and virtual tree rerouting and compare and contrast the performance of these two schemes under these rerouting schemes using simulation. Various performance metrics like the system throughput, total rerouting time, service disruption time, amount of buffering required at the MHs and the base stations are evaluated for comparative analysis using simulation. We also propose measures and mechanisms to improve the network performance using TCP under these rerouting schemes based on our analysis of the performance metrics.

The rest of the paper is organized as follows: in Section 2, we study the rerouting process. We provide a detailed view of the types of rerouting, and the handshake signaling involved in rerouting. We study the split connection TCP protocol and the snoop protocol in Section 3. Simulation is used to analyze the performance of the wireless TCP schemes under different rerouting techniques. The simulation set-up, our assumptions, the experiments conducted and the results are presented in Section 4. We analyze the results obtained from simulation in Section 5 and describe our conclusions and future work in Section 6.

Section snippets

Rerouting in mobile networks

In a mobile network, MHs move from one cell to another. When a MH moves from one cell to another, it registers with the base station of the new cell. Consider that there is currently a communication session between a source host and a destination host. If one of the hosts moves from its present cell, the session is interrupted. In order for the session to be restarted, a handoff needs to take place. Handoff is used by the mobile network to provide the MH freedom of motion beyond the cell

Wireless TCP schemes—split connection protocol and snoop protocol

As mentioned earlier, several schemes have been proposed by researchers to adapt TCP to wireless and mobile networks. Among the most popular of them are the split connection and the snoop protocol. In this section, we will take a detailed look at these schemes.

Simulation details

The simulation architecture is described in Fig. 12. The simulation was done using the PARSEC simulator developed by researchers at UCLA. PARSEC [5] is a discrete-event simulator that allows processes and systems to be modeled as entities that can interact with each other by message passing. We now describe the details of the simulation including the assumptions, network topology used, parameters used for simulation, the simulation experiments, performance metrics being evaluated and the

Analysis of results

We now analyze the performance of the various schemes based on the results of our simulation (see Fig. 13, Fig. 14, Fig. 15). In the case of source throughput and average round trip packet delay (Fig. 13, Fig. 14), we see that the snoop protocol TCP performs better the split connection TCP for all the rerouting schemes. This is because of faster re-transmissions by the snoop protocol TCP on the wireless link as compared to the split connection TCP. Also, the split connection TCP has the

Conclusion and future work

We have studied the performance of two important wireless TCP schemes—the split connection TCP and the snoop protocol TCP under full, partial, cell forwarding and virtual tree rerouting conditions. Snoop based virtual tree rerouting performs the best while split connection-based, cell forwarding performs the worst. Future work involves extending our work to other wireless retransmission schemes such as the link layer retransmission, explicit congestion notification and fast retransmissions.

References (29)

  • A. Acampora

    An architecture and methodology for mobile-executed handoff in cellular ATM networks

    IEEE Journal of Selected Areas in Communications

    (1994)
  • B. Akyol, D.C. Cox, Handling mobility in a wireless ATM network, Proceedings of IEEE INFOCOM'96, San Francisco, CA,...
  • B. Akyol, D.C. Cox, Rerouting for handoff in a wireless ATM network, Proceedings of IEEE ICUPC'96, Cambridge, MA, USA,...
  • E. Ayanoglu et al.

    Airmail: a link-layer protocol for wireless networks

    ACM/Baltzer Wireless Networks Journal

    (1995)
  • R. Bagrodia et al.

    Parsec: a parallel simulation environment for complex systems

    IEEE Computer

    (1998)
  • Ajay Bakre, B.R. Badrinath, I-TCP: indirect TCP for mobile hosts, Technical Report, Rutgers University, Technical...
  • H. Balakrishnan et al.

    Proceedings of ACM SIGCOMM

    (1996)
  • H. Balakrishnan et al.

    Proceedings of ACM Mobicom, Berkeley, CA

    (1995)
  • S. Bush, Handoff mechanism for mobile ATM systems, Internet Draft, http://tisl.ukans.edu/, Last Ref. 01/05/98,...
  • K.Y. Eng et al.

    A wireless a broadband ad-hoc ATM local-area network

    ACM/Baltzer Journal on Wireless Networks (WINET)

    (1995)
  • J. Ioannidis, D. Duchamp, G. Maguire, S. Deering, Protocol for supporting mobile ip hosts,...
  • R. Yuan et al.

    Proceedings of the Fifth Workshop on Third Generation Wireless Information Networks

    (1995)
  • R. Ghai et al.

    An architecture and communication protocol for picocellular networks

    IEEE Personal Communications

    (1994)
  • Mobile IP Working Group, Routing support for ip mobile hosts, Internet Draft,...
  • Cited by (1)

    • Interactive Transparent Networking: Protocol meta modeling based on EFSM

      2006, Computer Communications
      Citation Excerpt :

      We have particularly chosen two examples for demonstrating an implementation path via transparent networking. Please note because of their basic usefulness researchers have subsequently performed extensive performance evaluations [17,26] and have also proposed many other creative schemes [27,33]. The proposed transparency via interaction and the triggered TM deployment will provide them implementation paths as well.

    Preliminary version of this paper presented at the IEEE Region 10 International Conference on Global Connectivity in Energy, Computer, Communications and Control, New Delhi, India, 1998.

    View full text