CS625 - Advanced Computer Networks
Lecture 17 (12th September 2003 )
Scribe - Kumar Avijit
Roll # Y0170


Overview:

    1. Minimizing Latency
    2. Minimizing Size of packet
    1. Choices with the mobile host.
    2. Choices with the correspondent host.

The goal of mobile IP  is to enable the mobile host send and receive packets addressed with its home IP address, regardless of where the host resides. The idea is to separate address and location. This has been accomplished in mobile IP by means of having 2 addresses:
    1. The original IP address of the host.
    2. A care-of-address that is assigned to the host when it enters a foreign host. This address is assigned by the foreign host.

The basic model of the mobile IP infrastructure can be outlined by means of the following figure:

                                                                The Basic Mobile IP model

The bold arrow represents a tunneled IP packet. the normal arrows represent normal packets.  CH is the  correspondent host, HA is the home agent and MH is the mobile host. No tunneling is required if the MH is in the same administrative domain as the HA.
Tunneling is required if MH is inside a foreign network because an untunneled packet coming from HA to  MH bearing the source address of CH might be dropped by the border routers of the foreign network. Packets from MH to CH do not require any tunneling as MH can send the packets directly to CH without having the need to send them via HA.


Source Address Filtering and bidirectional tunneling:

The home agent encapsulates the correspoondent host's packets and correctly forwards them to the mobile host. The mobile host's replies will however fail to reach the CH if it uses the home address as the source address due to ingress filtering at the CH. To avoid this problem, the MH should use reverse tunneling to send replies to CH. This is called Bidirectional tunneling.
                                            Bidirectional Tunneling


Simultaneous Bindings:

A mobile host can register multiple bindings simultaneously. The home agent makes multiple copies of packets destined for the mobile host, and tunnels a copy to each care-of-address.

Simultaneous bindings may be used to :
Optimizations:

Consider the case when MH is close to the CH as in following figure:


                                    When CH is close to MH
In the above case, if the CH knows about its proximity with the MH, it can send packets directly to MH thus saving the extra distance of sending them via HA. Two ways have been proposed to make the CH smart enough to know of the care-of-address of the MH. The first is that when the HA forwards the packet to the MH, it may also send an ICMP message to the CH telling it of the care-of-address of MH. The second approach is an extension to DNS similar to the current MX records. The mobile agent, if it is not changing its location too frequently, can register its care-of-address with the extended DNS. The CH looks up the hostname and sees that  it has a care-of-address also. So it sends the packet directly to the care-of-address.

Tunneling a packet generally requires adding 20 bytes to the packet. If doing so makes the packet exceed the MTU, the packet needs to be fragmented, doubling the packet count. Thus encapsulation should be avoided wherever possible.

4 * 4 choices:
To achieve mobility transparency and successful delivery in security conscious networks, we use encapsulation. Though encapsulation increases size of packet, it has advantages also  as we can change the source address and destination address easily. The MH can then use the home IP address as the source address of the inner packet and the care-of address as the source address of the outer packet. There is another issue as to who should perform decapsulation. The MH can be sure of this service from its HA but sending each packet through the HA will sometimes lead to a considerable increase in the distance the packet travels through. If the MH sends the encapsulated packet directly to the CH, it will first have to make sure that the CH can decapsulate it.
Thus the MH has 4 choices to choose from in case of outgoing packets:
  1. Direct, encapsulated :In case of a decapsulation capable host. s = COA, d= D=CH. S=Home address
  2. Direct, unencapsulated: S=COA, D=CH
  3. Direct, unencapsulated: S=Home address, D=CH ( this will not work with security aware routers)
  4. Indirect, encapsulated: s=COA, d=Home agent, S=home address,D=CH
In the above choices, s and d are the source and destination addresses of the outer packet while their capital counterparts form the addresses in the inner packet.

Just as choices exist for the mobile host while sending packets to CH, similar choices exist for the corespondent host while sending packets to the MH.

  1. Indirect, encapsulated.(CH is unaware of mobility of MH) The packet will reach the Home network where it will be encapsulated by the HA before forwarding to MH.
  2. Direct, encapsulated: (CH aware of mobile host). d=COA,D=Home address, s=S=CH. In this case, the packet will be delivered directly to the MH.
  3. Direct, Home address: (This is feasible only if MH and CH are connected to the same network segment) S=CH, D=Home address.
  4. Direct, temporary address: S=CH, D=COA. In this case, the packet is delivered directly without encapsulation overhead. However sending packets this way forgoes the benefits of Mobile IP in that the  packets will be lost if the MH changes location.