Multicast Overview

Note: Both Protocol Independent Multicast (PIM) version 1 and PIM version 2 are supported. In this topic, the term PIM refers to both versions of the protocol.

Multicast traffic lies between the extremes of unicast (one source, one destination) and broadcast (one source, all destinations). Multicast is a “one source, many destinations” method of traffic distribution, meaning that the destinations needing to receive the information from a particular source receive the traffic stream.

IP network destinations (clients) do not often communicate directly with sources (servers), so the routers between source and destination must be able to determine the topology of the network from the unicast or multicast perspective to avoid routing traffic haphazardly. The multicast router must find multicast sources on the network, send out copies of packets on several interfaces, prevent routing loops, connect interested destinations with the proper source, and keep the flow of unwanted packets to a minimum. Standard multicast routing protocols provide most of these capabilities.

This chapter contains the following topics:

Multicast Architecture
Dense and Sparse Routing Modes
Strategies for Preventing Routing Loops
Multicast Protocol Building Blocks

Multicast Architecture

Multicast-capable routers replicate packets on the multicast network, which has exactly the same topology as the unicast network it is based on. Multicast routers use a multicast routing protocol to build a distribution tree that connects receivers (also called listeners) to sources.

Upstream and Downstream Interfaces

A single upstream interface on the router leads toward the source to receive multicast packets. The downstream interfaces on the router lead toward the receivers to transmit packets. A router can have as many downstream interfaces as it has logical interfaces, minus 1. To prevent looping, the router's upstream interface must never receive copies of its own downstream multicast packets.

Subnetwork Leaves and Branches

On a multicast router, each subnetwork of hosts that includes at least one interested receiver is a leaf on the multicast distribution tree (see Figure 35). The router must send out a copy of the IP multicast packet on each interface with a leaf. When a new leaf subnetwork joins the tree, a new branch is built so that the router can send out replicated packets on the interface. The number of leaves on an interface does not affect the router. The action is the same for one leaf or a hundred.

A branch that no longer has leaves is pruned from the distribution tree. No multicast packets are sent out on a router interface leading to an IP subnetwork with no interested hosts. Because packets are replicated only where the distribution tree branches, no link ever carries a duplicate flow of packets.

In IP multicast networks, traffic is delivered to multicast groups based on an IP multicast group address instead of a unicast destination address. The groups determine the location of the leaves, and the leaves determine the branches on the multicast network.

Figure 35: Multicast Elements in an IP Network

Image g003072.gif

Multicast IP Address Ranges

Multicast uses the Class D IP address range (224.0.0.0 through 239.255.255.255). Multicast addresses usually have a prefix length of /32, although other prefix lengths are allowed. Multicast addresses represent logical groupings of receivers and not physical collections of devices, and can appear only as the destination in an IP packet, never as the source address.

Notation for Multicast Forwarding States

The multicast forwarding state in a router is usually represented by one of the following notations:

(S,G) notation—S refers to the unicast IP address of the source for the multicast traffic and G refers to the particular multicast group IP address for which S is the source. All multicast packets sent from this source have S as the source address and G as the destination address.
(*, G) notation—The asterisk (*) is a wildcard for the address of any multicast application source sending to group G. For example, if two sources are originating exactly the same content for multicast group 224.1.1.2, a router can use (*, 224.1.1.2) to represent the state of a router forwarding traffic from both sources to the group.

Dense and Sparse Routing Modes

To keep packet replication to a minimum, multicast routing protocols use the two primary modes shown in Table 6.

Caution: A common multicast guideline is not to run dense mode on a WAN under any circumstances.

Table 6: Primary Multicast Routing Modes

Multicast Mode	Description	Appropriate Network for Use
Dense mode	Network is flooded with traffic on all possible branches, then pruned back as branches explicitly (by message) or implicitly (time-out silence) eliminate themselves.	LANs—Networks in which all possible subnets are likely to have at least one receiver.
Sparse mode	Network establishes and sends packets only on branches that have at least one leaf indicating (by message) a need for the traffic.	WANs—Network in which very few of the possible receivers require packets from this source.

Strategies for Preventing Routing Loops

Routing loops are disastrous in multicast networks because of the risk of repeatedly replicated packets, which can overwhelm a network. One of the complexities of modern multicast routing protocols is the need to avoid routing loops, packet by packet, much more rigorously than in unicast routing protocols. Three multicast strategies—reverse-path forwarding (RPF), shortest-path tree (SPT), and administrative scoping—help prevent routing loops by defining routing paths in different ways.

Reverse-Path Forwarding for Loop Prevention

The router's multicast forwarding state runs more logically based on the reverse path, from the receiver back to the root of the distribution tree. In RPF, every multicast packet received must pass an RPF check before it can be replicated or forwarded on any interface. When it receives a multicast packet on an interface, the router verifies that the source address in the multicast IP packet is the destination address for a unicast IP packet back to the source.

If the outgoing interface found in the unicast routing table is the same interface that the multicast packet was received on, the packet passes the RPF check. Multicast packets that fail the RPF check are dropped, because the incoming interface is not on the shortest path back to the source. Routers can build and maintain separate tables for RPF purposes. See Understanding PIM RPF Routing Tables.

Shortest-Path Tree for Loop Prevention

The distribution tree used for multicast is rooted at the source and is the shortest-path tree (SPT), but this path can be long if the source is at the periphery of the network. Providing a shared tree on the backbone as the distribution tree locates the multicast source more centrally in the network. Shared distribution trees with roots in the core network are created and maintained by a multicast router operating as a rendezvous point (RP), a feature of sparse mode multicast protocols.

Administrative Scoping for Loop Prevention

Scoping limits the routers and interfaces that can forward a multicast packet. Multicast scoping is administrative in the sense that a range of multicast addresses is reserved for scoping purposes, as described in RFC 2365, Administratively Scoped IP Multicast. Routers at the boundary must filter multicast packets and ensure that packets do not stray beyond the established limit.

Multicast Protocol Building Blocks

Multicast is not a single protocol, but a collection of protocols working together to form trees, prune branches, locate sources and groups, and prevent routing loops:

Distance Vector Multicast Routing Protocol (DVMRP) and Protocol Independent Multicast (PIM) operate between routers. PIM can operate in dense mode and sparse mode.
Three versions of the Internet Group Management Protocol (IGMP) run between receiver hosts and routers.
Several other routing mechanisms and protocols enhance multicast networks by providing useful functions not included in other protocols. These include the bootstrap router (BSR) mechanism, auto-rendezvous point (RP), Multicast Source Discovery Protocol (MSDP), Session Announcement Protocol (SAP), Session Description Protocol (SDP), and Pragmatic General Multicast (PGM) protocol.

Table 7 lists and summarizes these protocols.

Table 7: Multicast Protocol Building Blocks

Multicast Protocol	Description	Uses
DVMRP	Dense-mode-only protocol that uses the flood-and-prune or implicit join method to deliver traffic everywhere and then determine where the uninterested receivers are. DVRMP uses source-based distribution trees in the form (S,G) and builds its own multicast routing tables for RPF checks.	Not appropriate for large-scale Internet use.
PIM dense mode	Sends an implicit join message, so routers use the flood-and-prune method to deliver traffic everywhere and then determine where the uninterested receivers are. PIM dense mode uses source-based distribution trees in the form (S,G), and also supports sparse-dense mode, with mixed sparse and dense groups. Both PIM modes use unicast routing information for RPF checks.	Most promising multicast protocol in use for LANs.
PIM sparse Understanding IGMP and Multicastmode	Sends an explicit join message, so routers determine where the interested receivers are and send join messages upstream to their neighbors, building trees from receivers to an RP router, which is the initial source of multicast group traffic. PIM sparse mode builds distribution trees in the form (*,G), but migrates to an (S,G) source-based tree if that path is shorter than the path through the RP router for a particular multicast group's traffic. Both PIM modes use unicast routing information for RPF checks.	Most promising multicast protocol in use for WANs. See Understanding PIM and Static RPs.
PIM source-specific multicast (SSM)	Enhancement to PIM sparse mode that allows a client to receive multicast traffic directly from the source, without the help of an RP.	Used with IGMPv3 to create a shortest-path tree between receiver and source.
IGMPv1	The original protocol defined in RFC 1112, Host Extensions for IP Multicasting. IGMPv1 sends an explicit join message to the router, but uses a timeout to determine when hosts leave a group.	See Understanding IGMP and Multicast.
IGMPv2	Defined in RFC 2236, Internet Group Management Protocol, Version 2. Among other features, IGMPv2 adds an explicit leave message to the join message.	Used by default. ee Understanding IGMP and Multicast.
IGMPv3	Defined in RFC 3376, Internet Group Management Protocol, Version 3. Among other features, IGMPv3 optimizes support for a single source of content for a multicast group, or source-specific multicast (SSM).	Used with PIM SSM to create a shortest-path tree between receiver and source.
BSR and Auto-RP	Allow sparse-mode routing protocols to find RPs within the routing domain (autonomous system, or AS). RP addresses can also be statically configured.
MSDP Understanding SAP and SDP Multicast Session Announcements.	Allows groups located in one multicast routing domain to find RPs in other routing domains. MSDP is not used on an RP if all receivers and sources are located in the same routing domain.	Typically runs on the same router as PIM sparse mode RP. Not appropriate if all receivers and sources are located in the same routing domain.
SAP and SDP	Display multicast session names and correlate the names with multicast traffic. SDP is a session directory protocol that advertises multimedia conference sessions and communicates setup information to participants who want to join the session. A client commonly uses SDP to announce a conference session by periodically multicasting an announcement packet to a well-known multicast address and port using SAP.
PGM	Special protocol layer for multicast traffic that can be used between the IP layer and the multicast application to add reliability to multicast traffic. PGM allows a receiver to detect missing information in all cases and request replacement information if the receiver application requires it.