Detailed explanation of multicast routes in Sparse Mode

This article describes in detail the implementation of the Sparse Mode of multicast routing, mainly from the discovery of RP, the establishment process of the shared tree and source registration, the conversion from the shared tree to the source tree is described in detail.

I. RP discovery

RP discovery mechanism: Several routers are configured as alternative RP, and the alternative RP sends an interface containing its own campaign, which is generally used as the return port) to 224.0.0.39, the router configured as the RP er listens to this address, and selects the RP for comparing all selected routers, and notifies the IP address 224.0.0.40, all vrouters that enable multicast DENSE-MODE will listen to this address and learn the RP

2. Creation of the shared tree

1. The host sends the IGMP membership report to the group, and the leaf router adds *, G to the local routing table, and adds the interface linking the host to the output interface list,

2. Send the pim join message to the RPF interface. RPF points to the RP interface through the unicast route table)

3. For multi-channel access to the network, this task is done by the DR, And the sharing tree is built up through the DR. The message is continuously passed up until DR. After the shared tree is created, note that the sparse mode is an on-demand mode. No related shared tree will be generated on the Link without adding hosts.

4. if an interface in the forwarding status does not receive a group message from a downstream neighbor or host in three minutes, the interface will be deleted when the timer reaches 0. If a *, G) if the interface under the entry is blank, the *, G) Timeout timer will be started. If there are still no neighbors after three minutes, the *, G) entry will be deleted.

5. Shared Tree Pruning. The shared tree pruning is caused by the pruning message of the leaf router or downstream neighbor, which is roughly the same as that in dense mode.

6. To adapt to possible updates to the route table, the node performs RPF verification once every 5S.

Iii. Source Registration

1. The first hop router, after receiving the data stream from the source, creates S, G) locally. the outbound interface is empty, and the RPF interface is 0.0.0.0.

2. The first hop router uses unicast to encapsulate multicast data and send it to the RP in the registration message.

3. After receiving the RP, the RP knows that the source is alive. If the *, G) shared tree already exists, the registration message is unblocked and the original data is routed to the shared tree. Then, start to register the SPT with the source tree. The registration process is to send JOIN messages one by one to RPF (the direction is subject to the source) until the first hop router creates S, G) forwarding entries all the way, after the creation is successful, the first hop router begins to route the original data along the source tree multicast to the RP.

4. When the RP sends a reserved message unicast to the First-hop router, the first-hop router stops sending a unicast registration message containing multicast data.

5. Enable the time-out timer for the interface of the source tree S, G). If the interface does not receive the neighbor JOIN message within three minutes, it deletes the transfer-out interface. If the transfer-out interface is empty, (S, G) enable the timeout timer for the entry. If no JOIN message exists after the timeout timer expires for three minutes, the S, G) entry is deleted, and all S, G) entries are processed in this way.

4. Switch from the shared tree to the source tree SWITCHOVER)

1. when the leaf router receives the first data packet from the shared tree, it enables the conversion from the shared tree to the source tree by default). Otherwise, it will start when the value reaches the multicast data rate threshold.

2. if the RPF interface calculated by the RP and the RPF interface calculated by the source router are the same interface, only S, G) JOIN messages will be sent to the source RPF, start to build the Shortest Path Tree of the source tree ). If the RPF interface calculated by the RP and the RPF interface calculated by the source are not the same interface, the S, G) pruning message with the RP bit will be sent simultaneously, it will be passed up along the sharing tree. An S, G) entry will be created from the upstream until the RP, And the entry will be copied from *, G) to the interface, based on the calculation criterion of the RPF interface, the interface that receives the message is cut from the list of outgoing interface, RPF interface is calculated by RP, and a S, G) entry is generated. This ensures that the router that fails to reach the threshold can still receive messages correctly from the shared tree.

3. A very important principle is that when the routing table contains *, G) and S, G), data will be forwarded based on S, G. At the same time, this process triggers two routes. One is to register the source tree along the shortest path, and the other is to perform the RP-bit S, G to the sharing tree starting from the shard of the source tree and the sharing tree) the message is trimmed so that the related branches of the shared tree generate the S, G) forwarding entries that still direct RPF to the RP. However, the S, G) on the RP) an entry is generated after the source is registered. It only deletes the corresponding interface until the interface is cut.

4. The vro receives multicast data from the source tree SPT.