D-Link DI-1750 Reference Manual page 301

needn't wait for previous pruning states overtime, but send the engraft message to the previous actively
to make the pruning states return to transmit states.
Once source S remain send message to group G, the first hop router will periodically send (S,G)states
fresh information to next original broadcast tree to complete refurbish. The states fresh mechanism of
PIM-DM could fresh lower states, so pruning of broadcast tree ramus will not overtime.
In the multiple access network, except refer to DR voting, PIM-DM bring in following mechanism: use
assert mechanism to vote unique transmitor to prevent sending.
4. PIM-SM
PIM-SM (Protocol Independent Multicast Sparse Mode) is a multicast routing protocol of sparse mode.
In the network area running PIM-SM, all pim-sm-enabled routers will send hello messages periodically
to find the connected neighbors and select the Designated Router (DR) at the same time. A DR takes
charge of sending Prune/Graft messages towards the root of the multicast distribution tree for the
connected group members and forwarding the data packets from sources which are connected directly.
PIM-SM forms the distribution tree to forward data packets. There are two kinds of trees: the shared
tree rooted at the RP and the shortest path tree rooted at the source. And both kinds of trees are built
up and maintained by explicit prune/graft messages mechanism.
When a group member appears, it will express its interest of certain group through some membership
protocols just like IGMP. And on receiving this expression, the DR of this subnet will send (*, G) join
message of this group towards the RP. And with the join message traveling hop-by-hop towards the RP,
the shared tree is formed.
When a multicast data sender begins to send data destined for a group, the DR will encapsulate the
packet and unicast the data packets to the RP. This is so called registering. And if the RP is interesting
at the data, it will decapsulate it, forward it onto the shared tree and send (S, G) join message towards
the DR which have registered to him. If the RP is not interesting at the data or on receiving raw data
packets from the DR after sending (S, G) join to it, the RP will send register-stop message to the DR to
stop it from registering any more.
If a router is not interesting at the data of a group (or certain source of this group) any more, it will send
(*, G) or (S, G) prune message towards the RP.
In some cases, the RP tree is not the best path to forward the data from sender to some receivers.
Then the DR connected with the receiver can optionally initiates a transfer from the RP tree to the
source-specific shortest path tree when the data rate becomes larger than the threshold. And switch to
RP tree when the data rate becomes smaller.
Additionally, we adopt the BSR mechanism to determine the RP. So we need designate at least one
router to be the candidate BSR and one router to be the candidate RP. Then only one candidate BSR
will be elected to be the BSR. After the BSR elected, all the candidate RPs will advertise RP-candidate
messages to BSR. And the BSR will flood the RP-mapping information throughout this BSR area. Thus,
all the routers in this area will have the same RP-mapping information. According to it, and with the
proper hash mechanism, all routes will select the same RP for certain group. And you can also
configure the static RP manually.
5. DVMRP
DVMRP was the very first development of a multicast routing protocol. It was based on RIP and can be
viewed as a multicast version of the protocol. The purpose of RIP focuses on routing and forwarding
datagrams to a certain destination. This is also of DVMRP's concern. However, DVMRP also takes into
account the return paths back to the source.
By detemining the best path back to the source, a router would know the interface on which it has to
receive traffic from that source. In turn, this means that DVMRP calculates the best previous hop back
to the source network. RIP (Routing Information Protocol) differs in that it tries to find the best next hop
to the destination is calculated.
DVMRP uses Reverse Path Forwarding (RPF) to determine the best (shortest) path back to the source.
The router examines all the packets received as input to make sure that both source address and
interface are in the routing table. It looks up the source address in the forwarding table and compares
that entry with the receivng entry (RFP check). If the interface and entry do not match or are not in the
table, then the packets are discarded. If there is a match, then the router forwards the packets.
Model Name
- 299 -