Hi Linda, On 11/19/2012 07:11 PM, Linda Dunbar wrote: > Suresh, > > > >> -----Original Message----- >> >> Because it is :-). As an example, if you take an fat tree architecture, >> the number of access switch ports on a subnet is not directly related >> to >> the number of ports on the L3 gateway that are on the same subnet. >> > > For the following network, if one subnet (regardless IPv4 or IPv6) have hosts > in all server racks, the L2/L3 has to enable the subnet on links (or ports) > to all the access switches (ToRs). > > Then the IPv6's ND for the subnet will be flooded to all links. It is the > same problem as IPv4 ARP. How IPv6 ND scale better than IPv4's ARP?
At the risk of repeating myself. NO, the ND message will not be flooded to all the links. It is multicast to a solicited node multicast address (ff02::1:ffxx:xxxx) and the ND message is only sent to the links where there is at least one node that has joined this multicast group. > > > > > L2/L3 boundary > +-------+ > +/------+ | > | L2/L3 | \ > +---+---+/\ \ > / \ \ \ > / \ \ \ > +---+ +---+ +---+ +---+ > |T11|... |T1x| |T21| ... |T2y| ToR switches > +---+ +---+ +---+ +---+ > | | | | > +-|-+ +-|-+ +-|-+ +-|-+ > | |... | | | | ... | | > +---+ +---+ +---+ +---+ Server racks > | |... | | | | ... | | > +---+ +---+ +---+ +---+ > | |... | | | | ... | | > +---+ +---+ +---+ +---+ > > > p.s. I still don't see how it is related to implementation. May be you mean > network design? As I said above, if your switch implementation(s) do not support MLD snooping, then you will have an issue. In that case, the right fix is to add MLD snooping functionality into the switch. Thanks Suresh _______________________________________________ Int-area mailing list Int-area@ietf.org https://www.ietf.org/mailman/listinfo/int-area
