Possible Thesis Topics could be
---------------------------------
- Mobile IP Based Network Performance CComparison
- Performace validation of Mobile IP baased networks
- Performance for Wireless Network
- Effects of Mobility on the Transport Layer
Introduction
------------
When the name mobile computing comes into mind it means that people
are able to access network resources ubiquitously and seamlessly with
the help of either laptop computer, PDAs or electronic books. The
impact of mobile computing is wide spread. From tourism to field
research and from collaborative application to your garden, mobile
computing demonstrating its significance. The whole idea behind this
concept is the faster and reliable transfer of information without interruption.
This idea leads to such methods and protocols which allow network connectivity
across host movement. And to enable this mobility such protocol must not
require massive changes to devices involved in mobile computing.
Since the protocol for Internet is (transmission Control Protocol/Internet Protocol (TCP/IP), directly or indirectly most of the networks today are IP based networks. It means the protocl must be compatible with large installed base of IPv4 networks/hosts. IP is a network layer "best
effort" delivery protocol. It supports TCP and UDP protocols. An IP address consists of Network-ID and Host-ID. By design an IP address is tied to home network address, hosts are assumed to be immobile, intermediate routers only look at network address. If host is moved from its home network and attached to a new network wihout changing it IP address, then it won't be able to communicate with other hosts on the network. So the IP routing break under mobility.
To help eleviate problems just discussed, in 1996 IETF (Internet Enginering Task Force) proposed Mobile IP protocl. Mobile IP allows a mobile host to move about without changing its permanent IP address. Each mobile host has a home agent on its home network. Mobile host establishes a care
of address when it's away from home network.
Correspondant host is a host that want to send packets to mobile host. Correspondant host will always send packets to the mobile host's permanent IP address. These packets are routed to the mobile host home network. Home agent at home network intercepts the packets destined to the mobile host and forwards IP packets to current care-of-address. In response mobile host sends packets directly to the correspondant
The Transport Layer on the Mobile Communication Links
------------------------------------------------------
Traditional networks consist of wired links and stationary hosts, transport protocols support two basic type of communication: connection-oriented and connectionless. To use a connection-oriented transport protocol, two applications must establish a connection and then data travels across the connection. For example, the Transmission Control Protocol (TCP) provides a connection-oriented interface to applications. On the other hand, connectionless interface allows an application to send a message to any destination without prior connection establishment. For example, in the TCP/IP protocol suite, the User Datagram Protocol (UDP) provides connectionless transport.
TCP and UDP are widely used in the wired networks, and now mobile hosts and their wireless communication links are combined with wired networks to stablish future internetworks. Therefore, it is a very important subject how to apply TCP or UDP to the mobile communication links. However, communication over wireless links is characterized by limited bandwidth, high latencies, sporadic high bit-error rates, and temporary disconnection that network protocols and applications must deal with. These characteristics will degrade the performance of TCP or UDP in mobile communication links.
TCP performs very well on wired networks by adapting to end-to-end delays and packet losses. TCP provides reliability by maintaining a running average of estimated round-trip delay and mean deviation, and by retransmitting any packet, whose acknowledgement is not received within twice the deviation from average. Due to the relative low error rates over wired networks, packet losses are assumed to as a result of congestion. Nevertheless, in the presence of the high error rates present in mobile communication links, this assumption cause TCP to suffer a significant degradation in performance in the form of poor throughput and very high interactive delays when used in a wireless environment. This behavior arises due to the fact that mis-routed packet on the wireless link due to handoffs or errors on the wireless transmission due to harsh wireless links are correctly interpreted by TCP at the sender as congestion. In response to this
kind of "misunderstanding", TCP drops its transmission window size before retransmitting the packets, initiates the congestion control mechanisms such as slow start, and resets the retransmission timer using an exponential from the previous value, thereby unnecessarily reducing its bandwidth utilization.
H- Notable Factors for Mobility on the Transport Layer
H1-Loss of throughput
Mobile hosts (MHs) and mobility support routers (MSRs) collaborate to perform handoffs between cells. MSRs make their presence known by broadcasting periodic beacons over the wireless network. The motion with overlapping cells represents the best handoff performance possible. It is realizable in a real network only if overlap regions are large enough, and hosts move slowly enough, for handoff operations to complete while a moving host is still in the overlap region. The scenario with zero rendezvous delay represents the minimum network-level interruption introduced by non-overlapping cell handoffs. It is realizable only if the MH does not have to wait for a beacon before it can communicate with the new MSR, for example in a network where MSRs announce their presence by means of a continuous signal. Finally, the scenario with a 1-second rendezvous delay shows what happens as the length of network-level interruptions increases. It is a realistic scenario when a periodic beaconing scheme is used, since a MH may have to wait up to a full beaconing period before it receives a beacon from the new MSR.
H2-Pauses in communication
After the handoffs, there is typically a 1- to 3-second pause. During these pauses, TCP transmits no new data and transport-level communication comes to a halt. Also, the congestion window stops growing with every cell crossing. The congestion window is an upper bound on the transmission window, which controls how much unacknowledged data a TCP connection can have in transit over the network. Some time after the crossing, the windows shrinks to its minimum value and eventually begins to grow again.
H3-Packet losses
The long pauses in communication are caused by TCP's response to packet losses. Losses occur due to routing inconsistencies during non-overlapping cell handoffs. Consider the route from the MH to the fixed host (FH). When the MH leaves a cell without a warning, its routing tables continue to point to the old MSR. The MH does not know it has moved and therefore does not change its routing tables until a beacon arrives from the new MSR. Until then, the MH continues to send packets destined for the FH directly to the old MSR. These packets are lost because the MH can no longer reach the old MSR through the wireless interface.
Inconsistencies persist longer with the route from the FH to the MH. The old MSR does not know that the MH has left the cell until an explicit notification arrives from the new MSR, which cannot send the notification before it receives a greeting from the MH. Until the old MSR learns of the MH's motion, it continues to route packets directly to the MH. These packets are also lost because the old MSR can no longer reach the MH. Any other parts of the network involved in the handoff must also wait for explicit notification that the MH has moved before they can change their routing tables to point away from the old MSR to the new MSR.
H4-Slow Recovery
The congestion window drops abruptly after a cell crossing when the retransmission timer goes off, but returns only gradually to its previous level once transport-level communication resumes. TCP's slow start algorithm is responsible for this behavior. As acknowledgements reach the TCP transmitter, slow start first grows the congestion windows exponentially until it reaches one half of the window size at the time of the retransmission timeout, then grows it linearly. The slow start threshold thus decays exponentially with consecutive timeouts.
The slow recovery after each handoff contributes to the loss of throughput discussed before, but only moderately. Some measurements show that the algorithm throttles transmission for approximately 1 second after communication resumes. At that point the connection again reaches its maximum throughput and the congestion window stop affecting performance.
3.5 Unacceptable interactive response
Studies of human factor indicate that people perceive interactive response to be unacceptable if it takes longer than 100 to 200 msec. The halt of transport-level communication after non-overlapping handoffs discussed above could be 800 msec or longer. Also these pauses grow exponentially with growing rendezvous delays due to TCP's exponential retransmission backoff policy. Motion will thus lead to unacceptable interactive response unless using other methods.
Why did v chose TCP
-------------------
Reliable transport-layer protocols such as TCP are molded to perform well in wired networks in which packets losses can be assumed due to congestion of networks. Nevertheless, networks made up of wireless suffer from significant losses and hight bit-error-rate due to the harsh wireless links and handoff as well. TCP responds to all losses by invoking congestion control mechanism that results in degraded end-to-end performance in wireless networks.
Study of Simulation Tool for Mobile Communication Networks
-----------------------------------------------------------
It is difficult to conside suitable simulator for wireless and mobile communication that also supports mobile IP. Studies of at least 3 different kind of simulator namely OPNET, GloMoSIM and NS2 showed that NS2 fits well for this thesis work. NS2 supports mobile IP protocol and the significant part is that the mobile IP was tested on NS2 first before being proposed as an standar in IETF working group. NS network simulator version 2 is an event-driven network simulator embedded into the Tool Command Language, Tcl. An extensible simulation engine is implemented in C++ and is configured and controlled via a Tcl interface. A simulation is defined by a Tcl program. Once a network topology is defined and traffic sources and sinks are configured, the simulation is invoked and statistics are collected. A network topology is realized using three primitive building blocks: nodes, links, and agents. Nodes are passive objects which act as containers for agents, which are the objects that actively drive the simulation. Each node has
an automatically-generated and unique address. Traffic sources and sinks are examples of agents. After a topology is defined, agents are attached to nodes. Agents are the objects that actually produce and consume packets. Each agent is automatically assigned a port number unique across all agents on a given node, which is analogous to a TCP or UDP port. Links can be thought of as the physical transmission links, nodes as end hosts and routers, and agents as transport entities and/or processes that run on end host.
As stated above, the loss in mobile communication networks can be due to handoffs or transmission errors. One possible solution to handle transmission error is to employ the reliable link-layer protocols. These protocols attempt to hide link-related losses from the TCP sender by using local retransmissions in response to automatic repeat request (ARQ) and perhaps forward error correction (FEC) over the wireless link. However, because the LBL network simulator requires the well-defined topology before the simulation starts, it is impossible to simulate the handoff scenario, which needs to change the topology during the simulation.
Future Work
-----------
- UDP Protocol could be tested and the performance could be analyzed.
- Read time traffic using RTP protocol may be realized
- Wireless Mobile IP Network may be tessted for different type of TCP implementaion for instance REHNO, etc.
References
----------
-Mobile IP Presentation :
http://www.cs.uno.edu/~golden/mobile_ip/ppframe.htm
-List of Research papers published on ""TCP/IP over Air Links" topic
http://bbcr.uwaterloo.ca/~jpan/tcpair/
-Improving Performance for Wireless Nettwork in Transport Layer
http://www.ctr.columbia.edu/~wz23/e6950/report/projpaper2.html
Source code also included to simulate hand-off
(
geocities.com/shezy22/cisco_mip) (
geocities.com/shezy22)