Eliminating DAD / Optimistic DAD

By removing the DAD procedure altogether we can reduce the handover latency even further (potentially by a second or so). However, removing this check altogether is not a realistic option both in terms of ratification by the IETF or by tuning an implementation’s configuration. Thus, we are left with the option of fine tuning the DAD procedure in some way that reduces the time it takes for a MN to be able to use its CoA. A procedure called ‘Optimistic DAD’ which modifies [RFC2461] and [RFC2462] is proposed in [Moo05], which essentially allows a CoA to be used before it has completed DAD. The CoA is marked as ‘optimistic’ as opposed to ‘tentative’ before completing DAD and is marked as ‘preferred’ once DAD is complete.

Unfortunately, we have not been able to source a suitable implementation of Optimistic DAD with which to test. An implementation will soon be made available from Monash University, but this will appear too late for the lifetime of the 6NET project.

Conclusions

Our tests have demonstrated that even with fine tuning the parameters of routers for optimum MIPv6 handover performance, we still do not approach anywhere near good enough handover times for realtime voice/video applications.

We must therefore conclude that MIPv6, in its current form is not by itself sufficient to be the de-facto mobility management model in the mobile IPv6 Internet. Further optimisations relating to handover performance must be made in order to support interactive and real-time IPv6 applications in a mobile context.

We have examined the fast handover protocol for MIPv6, FMIPv6 [RFC4068]. This aims to improve handover latency by eliminating IPv6 configuration latency and also prevents packet loss by the use of a bi-directional tunnel while physical movement and MIPv6 CoA registration are taking place.

To the best of our knowledge no implementation has yet been developed for us to perform handover tests. Yet this does not prevent us from reasoning that FMIPv6 will indeed reduce handover latency in almost all cases. In some cases, e.g. predicted handovers and relatively infrequent movement FMIPv6 promises to be sufficient for the real-time applications, most notably the killer mobile application VoIP. However, without being able to perform real tests it would be rather hasty to take this for granted.