User's Manual

48-4
The working process of PIM-DM can be summarized as: Neighbor Discovery, Flooding & Prune, and Graft.
1. Neigh hour Discovery
After PIM-DM router is enabled, Hello message is required to discover neighbors. The network nodes which
run PIM-DM use Hello message to contact each other. PIM-DM Hello message is sent periodically.
2. Flooding & Prune of process
PIM-DM assumes all hosts on the network are ready to receive Multicast data. When some Multicast
Source begins to send data to a Multicast Group G, after receiving the Multicast packet, the router will make
RPF check first according to the Unicast table. If the check passes, the router will create a (S, G) table entry
and transmit the Multicast packet to all downstream PIM-DM nodes on the network (Flooding). If the RPF
check fails, i.e. the Multicast packet is input from the incorrect interface, and then the message is discarded.
After this procedure, in the PIM-DM Multicast domain, every node will create a (S, G) table entry. If there is
no Multicast group member in the downstream nodes, then a Prune message is sent to upstream nodes to
notify them not to transmit data of this Multicast group any more. After receiving Prune message, the
upstream nodes will delete the corresponding interface from the output interface list to which their Multicast
transmission table entry (S, G) corresponds. Thus a SPT（Shortest Path Tree, SPT） tree with source S as
root is created. The Prune process is initiated by leaf router first.
The process above is called Flooding & Prune process. Each pruned node also provides time-out
mechanics at the same time. When Prune is timed-out, the router will restart Flooding & Prune process.
The PIM-DM Flooding & Prune is periodically processed.
3. RPF Check
With RPF Check, PIM-DM makes use of existing Unicast routing table to establish a Multicast transmission
tree initiating from data source. When a Multicast packet arrives, the router will determine whether the
coming path is correct first. If the arrival interface is the interface connected to Multicast source indicated by
Unicast routing, then this Multicast packet is considered to be from the correct path. Otherwise the Multicast
packet is to be discarded as redundant message. The Unicast routing message used as path judgment can
root in any Unicast Routing Protocol, such as messages found by RIP, OSPF, etc. It doesn’t rely on any
specific Unicast Routing Protocol.
4. Assert Mechanism
If each of two Multicast routers A and B on the same LAN segment has a receiving route respectively and
both will transmit the Multicast packet to the LAN after receiving the Multicast data packet sent by the
Multicast Source S, then the downstream node Multicast router C will receive two exactly same Multicast
packets. The router needs to choose a unique transmitter through Assert mechanism after it detects this
situation. An optimal transmission path is selected through sending out Assert packet. If the priority and
cost of two or more path are same, then the node with larger IP address is taken as the upstream neighbor
of the (S, G) entry and in charge of the transmission of the (S, G) Multicast packet.
5. Graft
When the pruned downstream node needs to recover to transmission status, this node uses Graft Packet to
notify upstream nodes to restore multicast data transmission.