The Neighbor Discovery Protocol(NDP) operates in the link-layer and is responsible for the discovery of other nodes on the link. It determines the link-layer addresses of other nodes, finds the available routers, and maintains reachability information about the paths to other active neighbor nodes.

Router Advertisement (RA) is one of the IPv6 Neighbor Discovery Protocol (NDP) packets that is used by the hosts to discover available routers, acquire the network prefix to generate the IPv6 addresses, link MTU, and so on. The routers send RA on a regular basis, or in response to hosts Router Solicitation messages.

IPv6 wireless client mobility manages the IPv6 RA packet . The converged access switch forwards the link-local all-nodes multicast RA packets to the local and roaming wireless nodes mapped on same VLAN the RA was received on.

Figure 1 illustrates the link-local all-nodes mcast RA forwarding issue in the wireless node mobility.

Figure 1. Roaming Client Receiving Invalid RA from Router 2

Figure 2 illustrates how a roaming client “MN” receives RA from VLAN 200 in a foreign switch and how it acquires an new IP address and breaks into L3 mobility's point of presence.

Figure 2. Roaming Client Receives Valid RA from Router 1

To safeguard the power-saving wireless clients form being disturbed by frequent unsolicited periodic RAs, the controller can throttle the unsolicited multicast RA.

There are three ways for IPv6 client to acquire IPv6 addresses:

• Stateless Address Auto-Configuration (SLAAC)
• Stateful DHCPv6
• Static configuration

For these methods, the IPv6 client always sends NS DAD (duplicate address detection) to ensure that there is no duplicated IP address on the network. The switch snoops the clients NDP and DHCPv6 packets to learn about its client IP addresses and then updates the controllers database. The database then informs the controller for the clients new IP address.

In the case when the new IP address is received after RUN state, whether an addition or removal, the controller updates the new IP addresses on its local database for display purposes. Essentially, the IPv6 uses the existing or same PEM state machine code flow as in IPv4. When the IP addresses are requested by external entities, for example, from Prime Infrastructure, the controller will include all the available IP addresses, IPv4 and IPv6, in the API/SPI interface to the external entities.

An IPv6 client can acquire multiple IP addresses from stack for different purposes. For example, a link-local address for link local traffic, and a routable unique local or global address.

When the client is in the DHCP request state and the controller receives the first IP address notification from the database for either an IPv4 or IPv6 address, the PEM moves the client into the RUN state.

When a new IP address is received after the RUN state, either for addition or removal, the controller updates the new IP addresses on its local database for display purposes.

When the IP addresses are requested by external entities, for example, from Prime Infrastructure, the controller provides the available IP addresses, both IPv4 and IPv6, to the external entities.

The switch supports IPv6 client as seamlessly as the IPv4 clients. The administrator must manually configure the Vlans to enable the IPV6, IPv6's snooping and throttling functionality. This will enable the NDP packets to throttle between the switch and its various clients

The switch will sync with the wireless clients when the clients IP address is hard to learn. When a switchover happens, the IPv6 neighbor binding table is synced to standby state. However, the wireless client will itself disassociate and reassociate to a new active state once the switchover is complete and the neighbor binding table is updated with latest information for that client.

If, during the reassociation, the client moves to another AP then the original entry in the binding table is marked as down for sometime and will be aged-out.

For the new entries joining the switch from another AP, the new IP address is learned and notified to the controller's database.

Note	This feature is available only for the Cisco Catalyst 3850 Switch.