![[previous]](../../../graphics/l_arrow.gif){kind=icon}
![[next]](../../../graphics/r_arrow.gif){kind=icon}
This chapter describes the IP multicast routing implementation on the CoreBuilder 2500 system. It includes these sections:
IP multicast routing is an extension of the Internet Protocol. With multicast routing, a router or switch sends packets to a specific group of hosts without using broadcasts or multiple unicast transmissions. Multicast destinations include:
Multicast routing operates without loops or excess transmissions.
The CoreBuilder 2500 system supports two IP multicast protocols:
This chapter describes these protocols and the algorithms that the CoreBuilder 2500 system uses for multicast routing.
The CoreBuilder 2500 system is capable of dynamic multicast filtering based on the Internet Group Management Protocol (IGMP). This protocol ensures that multicast packets are flooded only to appropriate ports in a routing interface.
IGMP tracks end station group membership within a multicast group. Membership in a group is dynamic, and hosts can be a member of more than one group at a time. The system avoids propagating multicast broadcasts to the entire subnetwork by confining them within the multicast group (a process called IGMP snooping).
The Distance Vector Multicast Routing Protocol (DVMRP) establishes multicast delivery paths over a series of routing devices. DVMRP is a distance-vector-routing protocol, similar to the IP Routing Information Protocol (RIP). Multicast routers exchange distance vector updates that contain lists of destinations and the distance in router hops to each destination. They maintain this information in a routing table.
The Internet Multicast Backbone (MBONE) uses DVMRP. Because of DVMRP, the CoreBuilder 2500 system can establish delivery paths without direct connections to multicast routers.
The MBONE is the Internet's experimental multicast backbone network. Users can test multicast applications and technology on the MBONE without waiting for Internet multicast standards to be set. You can connect to the MBONE through any Internet service provider (ISP).
MBONE routers forward multicast packets over an interface or over a multicast tunnel only if the Time-To-Live (TTL) value in the packet is larger than the tunnel's threshold. See "Multicast Tunnels" on page 5-5 for more information about tunnels.
At software revisions earlier than 8.0, CoreBuilder 2500 systems that are connected to the MBONE network support up to 16 IP multicast tunnels or routing interfaces. CoreBuilder 2500 systems at revision 8.0 or later support up to 32 IP multicast tunnels or routing interfaces connected to the MBONE.
The CoreBuilder 2500 system uses three multicast routing algorithms:
In most flooding algorithms, a network node receives a packet that was sent to a multicast destination. The node determines whether the packet is an original that it has not received before or a duplicate of a packet that it has received before. If the packet is original, the node forwards the packet on all interfaces except the incoming interface. If the packet is a duplicate, the node discards it.
This flooding algorithm is useful when network robustness is important. The algorithm does not depend on routing tables. Multicast destinations receive packets as long as at least one path to the destinations exists and no errors occur during transmission.
The Spanning Tree algorithm detects loops and logically blocks redundant paths within the network. The paths form a loopless graph, or tree, spanning all the nodes in the network. A port in the Spanning Tree blocking state does not forward or receive data packets.
After the algorithm eliminates redundant paths, the network configuration stabilizes. When one or more of the paths in the stable topology fail, the protocol automatically recognizes the changed configuration and activates redundant links. This strategy ensures that all nodes remain connected.
Figure 5-1 shows a simple network with six links.
Figure 5-1 Simple Network Implemented Without Using Spanning Tree
Figure 5-2 shows a spanning tree for the network in Figure 5-1. The spanning tree consists of links 1, 2, 3, and 4.
Figure 5-2 Spanning Tree Algorithm Implemented to Block Redundant Paths
The MBONE network uses the reverse path forwarding (RPF) multicast algorithm. RPF avoids duplicate paths on multiaccess links by using a routing table to compute a logical spanning tree for each network source. The algorithm consists of three steps:
1 . When the system receives a multicast packet, the algorithm notes the packet's source network and the interface that received the packet.
2 . If the receiving interface is on the shortest path toward the source network, the system forwards the packet to all interfaces except the interface where the packet was received.
3 . If the receiving interface is not on the shortest path toward the source network, the system drops the packet.
Pruning is a method used in the RPF algorithm to forward packets to a spanning tree only if group members exist in the tree. This method results in fewer spanning trees, but it requires dynamic updates to the routing table.
Nodes that are at the border of the network and have no point beyond them in the RPF spanning tree are called leaf nodes. Leaf nodes all receive the first multicast packet. If a group member is attached to a leaf node, the node continues to accept packets. If no group member is attached to the leaf node, the node sends back a prune message to the router that sent the packet. The message notifies the router not to send any further packets to this group. In the CoreBuilder 2500 system, the Administration Console IP multicast CacheDisplay includes information about when pruning occurs on the Spanning Tree.
This section describes the characteristics of the CoreBuilder 2500 multicast interface.
The DVMRP metric value determines the cost of a multicast interface. A higher cost results in a slower link. The default value is 1.
The TTL threshold determines whether the interface forwards multicast packets to other switches and routers in the subnetwork. If the interface TTL is greater than the packet TTL, then the interface does not forward the packet. The default value of 1 means that the interface forwards most packets.
The rate limit determines how many multicast packets can travel over the interface per second. The CoreBuilder 2500 system drops multicast traffic that travels faster than this rate. The default value of 0 means that no rate limit is in effect. A lower rate limit results in fewer multicast packets traveling over the interface.
Multicast tunnels logically connect two multicast routers through one or more unicast routers. The multicast router at the local endpoint of the tunnel puts multicast packets in a format that unicast routers can interpret and forward. The multicast router at the receiving endpoint reformats the packets into their multicast format. Tunnels are virtual links through the unicast IP network.
Multicast tunnels have characteristics similar to those of a multicast interface: a DVMRP metric value, a TTL threshold, and a rate limit. When you define a multicast tunnel, you also specify the destination address of the multicast router at the tunnel's remote endpoint.