10/11/2000  Mobile IP

Scribed by B. Hoon Kang.

We discussed three papers about mobile IP, its route optimization (IMHP) and mobility management in PCN.

The followings are the material not on the slides but discussed during the class.

As an alternative to Mobile IP section we detailed 

Loose Source Routing :

A-> B->C->D , where A want to send a packet to D.  Using LSR, each router adds their path into the IP packet so that the path(routers) that has been forwarded/traversed until destination can be recorded. And then the destination can send the packet which can be forwarded in reverse order later.   

A 

B

C

D

Obviously, it has a security problem since the packet can spoof that it comes from the secure source and routes.  In other words, it can make the source address to be appeared to be secure source so that it can pass firewall.

The 3rd problem with LSR  that is not on the slide is that the LSR requires recomputation of checksum since the packet is being added with route information and the checksum of the IP packet has to be re-computed whenever the route info. is added to the packet, which is a performance hit!!.

(More detailed discussion beyond class material can be found in the paper : Mobile Host Internetworking Using IP Loose Source Routing (1993)  (Correct)  (13 citations) David B. Johnson February 1993 CMU-CS-93-128 School of Computer Science...) Note inserted by the scriber.

Anthony mentioned that the mobility are useful for long-lived sessions such as SSH, Voice over IP, and the slow (or not at all) deployment of Mobile IP could be due to the fact that there is not much of needs for the mobility support since most other application such as email, web does not require mobility support.

In discussing the Mobile IP, we lengthened the proxy ARP where Home Agent proxy ARPs for its MH (Mobile Host) (by  maintaining the current care-of address of its Mobile Host ).

Proxy ARP

Router maintains the following table where IP address maps into MAC address since MAC is easier to filter out the packet at the receiver end.  For example, the network card at 128.131.4.2 has MAC address 0x954924(48bit) and when the router gets a packet destined to 128.131.4.2 then it will look up its MAC address in its ARP Table (below diagram) if it is not there then the Router sends out the ARP (Address Resolution Protocol) packet saying I have packet for 128.131.4.5 what is MAC for this and who has it ?  Then, the destination reply with its MAC address saying I have it.  The router then update its ARP table and sends the packet to the destination MAC address.  The destination's network card can be configured to listen only on the packet sent to its MAC address, which is more efficient than the promiscuous mode that has to listen on every packet.

In the authentication of the MH's new care-of address report to HA, they use nonce to avoid replay attack. And the nonce is better than timestamp in that synchronizing clock in a distributed environment is hard to manage and also has a skew time which could be exploited.

We discussed the Mobile IP's tunneling where the packet form CH has to go through HA -> FA -> MH.  And the IP-to-IP encapsulation requires extra 20 bytes which is more than 20% or 100% of IP packet whose size typically 128byte or 20 bytes for voice.

To solve this triangulation problem, we discussed the IMHP optimization where the binding info(as cache) are maintained at any IMHP aware CH so that it can directly tunneling to the MH's care-of address rather than going through HA.  MH notify its current location to the CH only when they try to send a packet, sort of Lazy Binding.

Anthony also mentioned the binding notification backoff (like binary backoff in congestion control). If the MH sits stationary it backs off the rate of sending out the binding notification otherwise a very stationary MH has to send out the same binding notification meaninglessly consuming computations.

After 5 minute break, we looked at the PCN's mobility scheme, IS-41 for AMPS and GSM's mobility part, which is widely used than Mobile IP, maybe due to its demands in voice traffic or its centralized update method and agreement among providers.

Most of materials are on the slides, additionally Anthony mentioned the problem of PCN's mobility in scaling.  Back then there was not many providers hence the roaming agreement was not an issue but nowadays there is more than 300 providers and managing the billing,VLR/HLR updates according to different 300 more roaming agreement should have more scalable solution than the current PCN's mobilty scheme.  Moreover, SS#7 has no security since it was designed to be used in private network, which is not easy to be enforced with growing number of different providers.  He also mentioned a tradeoff in assigning TDLN (Temporay Local Directory Number) which is assigned when the MH enters new VLR and making calls.  If TDLN is assigned whenever MH enters new VLR as long as your phone is turned on, you might have to pay for the service charge without making any phone calls by simply passing through new roaming zone with your MH device turned on.  If TDLN is assigned only when the user is trying to make a call, then the user might have longer delay in making calls, which is a tradeoff.  

We also discussed the payment issue where HA gives 3 challenge secret to VA and VA challenges the MH periodically or once per call which is up to the agreement between providers.  For example, if calls get over 10 minute the VA has to perform challenge response involving with HA if the agreement requires 1 challenge response in every 10 minute.

When we discuss the optimization of mobility location management, ( the location management traffic for mobility is non trivial), Anthony mentioned the Highway is a good example to project (guess) the trajectory of moving host(MH).