Hi Pete, > I really don't think we should force DHCP to re-run on every handover. >I thought one of the motivations for network-based mobility was to >minimize the signaling to the MN during handover.
In the current mobility models we have two DHCP deployment models. a.) DHCP server in the access. b.) DHCP Relay in the access For #a, it is the dominant model. The DHCP server is collocated on the access gateway; the interworking between the DHCP server and the mobility function in the access gateway allows the network to offer the same IP address to the mobile node. After each handoff and based on DNA considerations, the MN may go into DHCP INIT-REBOOT followed by Request/reply. I don't know if I should call this as complete Re-Run after each handoff, but at least the current data suggests the handoff latency related to this same-link validation is insignificant. Now, from coloring perspective, the DHCP server in the access network can potentially change the properties of the address. MN is on a new world, new access network, new link and hence some changes to the properties. For #b, we continue to route the DHCP lease renewal messages back to the network where the MN obtained its address. Typically its tunneled to the anchor. Here I see your point of not impacting the DHCP state machine by changing properties based on MN's movement. But, if the DHCP server is aware of changes to properties, can it not send the updated properties ? We have to look at properties as meta-data that goes with an IP address/prefix. This meta-data should not have any relation to the DHCP state machine. But, some properties of that address do change, based on MN's movement, mobility state changes ..etc. If the DHCP server is aware of these property change, IMO, it should reflect the updated properties. I'm more interested in #a and not deal with this issue at all. But, if we insert a property element in PIO, we should do that consistently and have that in DHCP as well. Regards Sri On 3/20/14 5:37 PM, "Peter McCann" <[email protected]> wrote: >Hi, Sri, > >Sri Gundavelli (sgundave) wrote: >> >> >> >> >> On 3/20/14 5:54 AM, "Peter McCann" <[email protected]> wrote: >> >>> I think our extensions should be to the prefix information option and >>> not DHCP. >>> >>> The properties of an address may change after a handover and we should >>> not couple the DHCP state machine (which is about lease renewal) to the >>> handover state machine. >> >> Hi Pete, >> >> This is a good comment. If we are making any assumption that even >> after a handover, the DHCP transactions are still hitting the same >> DHCP server node, this may be a valid concern. But, if DHCP >> transactions are locally terminated after handover (Ex: RFC5844/MAG), >> then the updated properties can be provided as part of the new DHCP >> transaction. Some parameters such as MTU settings do change after an >> handover and so we can probably agree that address properties can >>change as well. >> >> >> Regards >> Sri > >I really don't think we should force DHCP to re-run on every handover. I >thought >one of the motivations for network-based mobility was to minimize the >signaling >to the MN during handover. > >-Pete > > _______________________________________________ dmm mailing list [email protected] https://www.ietf.org/mailman/listinfo/dmm
