
IP Multicasting 11
Operation of PIM Sparse Mode
Once roles are established, multicast routing follows specific phases:
1. Rendezvous point tree
2. Register stop
3. Shortest path tree
While multicast routing always begins with phase 1, the designated router for a
receiver determines whether and when to move on to phases 2 and 3,
depending on the amount of traffic from the source.
Rendezvous point
tree
Phase 1 establishes and uses a shared tree rooted at the rendezvous point (RP)
to forward all multicast data to group members.
When an IP host sends an IGMP join message to the local PIM designated
router, which is not the RP for the group, the designated router sends a PIM
join message towards the RP for the group (“upstream”). The designated router
determines which router is the RP for the group from the most recent bootstrap
message. Every router the join message passes through records that there is a
group member on the incoming interface. Eventually, the join message reaches
either the RP, or another router that already knows that it has a group member
downstream. If the group has many members, the join messages converge on
the RP to form a rendezvous point tree (RPT). This is called a shared tree
because multicast data that is sent to the group by any sender shares the tree.
The multicast receiver’s designated router sends join messages periodically
according to the upstream join timer as long as the IP host is a member of the
group. When the last receiver on a subnetwork leaves the group, the join
messages stop, and their entries timeout on routers that are closer to the RP.
The sender’s designated router encapsulates the multicast data in a unicast
packet in a process called registering, and sends these register packets to the
group’s RP. When the RP receives the data, it decapsulates them, and forwards
them onto the shared tree.
Register stop Phase 2 improves efficiency and performance by using register stop. In this
phase the RP joins the shortest path tree between the source and receiver. This
allows the original (unencapsulated) packets to be forwarded from the sender,
instead of encapsulated packets. It also allows shorter paths to receivers that
are close to the sender, making it more efficient in some circumstances.
When the RP for a group receives the first encapsulated data packet from a
source, it joins the shortest path tree towards the sender. Once data is able to
flow along the shortest path from the sender to the RP, packets do not need to
be registered. The RP sends a register stop message in reply to the next
encapsulated message. When the sender’s DR receives the register stop
message, it stops registering. The DR sends a null register message to the RP to
find whether the RP still does not need to receive registered packets. If it
receives another register stop message, the DR continues to forward only the
native data packets. If the DR does not receive another register stop message
within the register probe time, it resumes registering the data packets and
sending them to the RP.
When the RP starts receiving native data packets from the source, it starts to
discard the encapsulated packets, and starts forwarding native packets on the
shared tree to all the group members. If the path from the source to the RP
intersects the shared RP tree for the group, then the packets also take a short-
cut onto the shared tree for delivery to the group members down its branches.
Komentáře k této Příručce