We have tested manually configured IPSec security associations between Microsoft MN and KAME HA. This seems to work well in that the Microsoft MN is able to register with the KAME HA and binding updates are performed successfully. When testing a Linux MN or HA we have only used non- IPSec authenticated communication between the MN and HA. The same is true when using a Cisco HA. In general we have found that all the implementations in our testbed are interoperable in a basic form.

For obvious reasons we do not employ any AAA or access control mechanisms when performing handover latency tests. This is especially important when using any streaming applications as the delay incurred when waiting for network access to be granted can result in a considerable number of lost packets. We have found that using IPerf [IPERF] between the CN and MN is a very useful tool for discovering the extent of packet loss during handovers when comparing different network and different tunings of the RA intervals.

we configure each Home Agent to announce unsolicited Router Advertisements at more frequent intervals than the default. In general, we use a Router Advertisement interval of 1 second which is sufficient for most adaptable TCP applications such as email, web, ftp etc. However, we have noticed considerable (i.e. unpleasant) disruption when using a 1 second interval when streaming video and/or audio. Reducing the interval to around 300ms seems to be fair benchmark to set for supporting streaming media. Although it is possible to reduce the interval even further, in most tests we did not observe enough improvement in the stream disruption to justify the extra link bandwidth and Home Agent CPU cycles being used.

Initially, the most simple way of testing the operation of the MIPv6 testbed is by testing basic connectivity between the Mobile Node and Correspondent Node with a stream of ping requests from the Correspondent Node to the Mobile Mode while the Mobile Node moves between its home network and a foreign network. In a correctly configured Mobile IPv6 testbed, the ping requests should continue to be answered as the Mobile Node moves. Sometimes we observe one ping timing out as the packet is lost. This can occur if the ping request is sent just as the Mobile Node becomes unreachable in its original location but has not yet completed the binding update procedure to be reachable at its new location. However, on a properly configured testbed we rarely observe more than one ping timeout.

For testing that TCP connections survive movement of the Mobile Node, a simple test is to run telent or similar (using the client and server in either combination of Mobile Node and Correspondent Node) and observe that the session remains active after movement. You should not observe any noticeable effects (save some possible interruption in character echo on the terminal) if all is configured correctly.

Beyond these basic tests some users may want to experiment with the TAHI compliance test suite described at [TAHI].