Imagine streaming live video to thousands of viewers simultaneously. How can we do this without overwhelming the network? The answer is IP multicast.

Traditionally, networks relied on unicast (one-to-one) and broadcast (one-to-all) communication. Broadcast, while useful for reaching multiple devices, floods the network, causing unnecessary overhead and security concerns. IP multicast provides a more efficient solution. In IP multicast, a sender transmits data to a specific group identified by a multicast address within the range 224.0.0.0 to 239.255.255.255 (Class D). Devices that want to receive this data must then 'subscribe' or join that specific multicast group. This ensures that only interested devices receive the multicast traffic.

IP multicast is used in a variety of applications, from IPTV and Hospital TV to network protocols like VxLAN, First Hop Redundancy Protocols, and routing protocols like BGP and OSPF. In this blog, we will delve into the mechanisms behind IP multicast, exploring how devices join groups, how routers forward multicast traffic, and the different multicast modes.

The Process of IP Multicast

The process of IP Multicast can be divided into two main parts:

1. Joining a Multicast Group: Receiver-to-Router Communication: Receivers use group membership protocols to inform their local router of their interest in receiving traffic for a specific multicast group.
2. Distributing Multicast Data: Sender-to-Router Communication (Distribution Trees): Routers build distribution trees to create efficient and non-redundant paths for multicast data to reach all interested receivers.

Joining a Multicast Group: Receiver-to-Router Communication

Receivers express interest in multicast groups using group membership protocols. In IPv4 networks, this is done with the Internet Group Management Protocol (IGMP), while IPv6 networks use Multicast Listener Discovery (MLD).

IGMP (IPv4)

• IGMPv2: This version uses ASM (Any Source Multicast) to signal its interest in groups. With ASM, the receiver is only able to specify the interested group, but it cannot specify the exact source of the intended group.
• IGMPv3: This version uses SSM (Source-Specific Multicast) to signal its interest in groups. Receivers in such deployments have the ability to specify the source of the intended group for subscription.

MLD (IPv6)

• MLDv1: The IPv6 counterpart for IGMPv2. It uses ASM deployment.
• MLDv2: The IPv6 counterpart for IGMPv3. It uses SSM deployment.

What versions of these protocols are used on Cisco Nexus Switches by Default?

• IGMPv2 (ASM) is the default for IPv4 on Nexus switches. This is due to the historical context and its initial simplicity.
• MLDv2 (SSM) is the default for IPv6 on Nexus switches. MLD benefited from the lessons learned from IGMP, recognizing the advantages of SSM for scalability and security.

To use IGMPv3, all participating interfaces must be configured for it. This is crucial. IGMP versions on a network segment must be consistent. You can't mix and match.

How do switches know not to send IGMP traffic to all switches in the network?

While switches don't inherently participate in IGMP (Internet Group Management Protocol) like routers, they need to be aware of IGMP messages to efficiently manage multicast traffic within a local network. They achieve this through IGMP snooping, where the switch "listens" to IGMP communication between hosts and routers on its ports. By examining IGMP queries, membership reports, and leave messages, the switch builds a table that maps multicast groups to the ports that have expressed interest in those groups. When the switch receives multicast traffic, it uses this table to filter and forward the traffic only to the interested ports, preventing unnecessary flooding and bandwidth consumption.

Distributing Multicast Data: Sender-to-Router Communication (Distribution Trees)

When a router receives signals from receivers about a multicast group, it uses Protocol Independent Multicast (PIM) to locate the sender. PIM is protocol-independent, working with both IGMP and MLD.

What are the types of Distribution Trees PIM creates?

PIM uses distribution trees for efficient, non-redundant paths between sender and receiver. Two types exist:

• Source Distribution Tree: Builds a shortest path tree (SPT) between sender and receiver. Creates an SPT for every group and sender, which is optimal but memory-intensive.
• Shared Distribution Tree: Routers agree on a rendezvous point (RP) to which multicast are sent. The RP then forwards to the receiver. An SPT is created per group, not per sender. The RP concept is similar to a Designated Router in OSPF or a Route Reflector in BGP.

How does PIM handle traffic?

• Sparse Mode (PIM-SM): PIM-SM uses a "pull" method, sending multicast traffic only to routers with interested receivers. This is more scalable and avoids network flooding. PIM-SM typically uses a Shared Distribution Tree (RPT) initially. Receivers send "join" messages towards the Rendezvous Point (RP). Senders register with the RP. When a sender sends traffic for a group, the RP forwards it to interested receivers. PIM-SM can also switch to a Source Distribution Tree (SPT) for more direct forwarding.
• Dense Mode (PIM-DM): PIM-DM floods all routers in the domain with multicast traffic. Routers without interested receivers downstream "prune" (stop) the traffic flow. This method is less scalable due to the initial flooding and is generally only practical in small networks.

What are the types of IP Multicast Routing Protocols used for routing Multicast?

There are two major types of multicast routing protocols:

Interdomain Multicast Routing

This enables multicast communication between multiple domains or autonomous systems (ASes). Key protocols include:

• MSDP (Multicast Source Discovery Protocol): MSDP is used to announce the availability of multicast sources between rendezvous points (RPs) in different domains. The Designated Router (DR) on the source's network segment sends information about the source to its local RP. The RP then uses MSDP to share this information with RPs in other domains.
• MBGP (Multiprotocol BGP): MBGP extends BGP to carry multicast routing information in addition to unicast routes. It enables the exchange of IP prefixes that are used for multicast Reverse Path Forwarding (RPF) checks between ASes. MBGP doesn't forward the multicast data itself; it provides the necessary routing information to multicast forwarding protocols (like PIM) so that routers in each AS can build inter-AS distribution trees.

Intradomain Multicast Routing

These protocols demonstrate how IP Multicast routing is done in a domain. All intra-domain multicast routing makes use of PIM, but in different modes. They include:

• PIM-ASM: PIM-Any Source Mode makes use of Shared Distribution Trees for multicast distribution. It uses the RP to discover sources and receivers, but also has the optionality of forming source-to-receiver SPTs directly as a form of optimization. With RIM-ASM, the receiver does not specify which source from the multicast group, and that is why RP can be used as it would allow for the RP to pick any source it has recorded as a sender for that group.
• PIM-SSM: PIM-Specific Source Mode utilizes Source Distribution Trees, specifically Source-Specific Trees (SPTs), for multicast distribution. The key characteristic of PIM-SSM is that the receiver explicitly knows which source it wants to receive multicast traffic from within a particular multicast group. This knowledge is crucial. PIM then leverages the underlying IGP (Interior Gateway Protocol) information to build the necessary SPTs. These SPTs create the shortest path from the specific source to the interested receiver(s). Because the receiver knows the source, there's no need for shared trees or rendezvous points (RPs), simplifying the protocol and improving efficiency.
• BIDIR-PIM: Bidirectional-PIM is used for many-to-many multicast relationships. While BIDIR-PIM typically uses a Rendezvous Point (RP), it's not an absolute requirement like in PIM-SM (ASM). A key difference from ASM is that in BIDIR-PIM, a node can be both a receiver and a sender. This is essential for many-to-many communication. This approach is beneficial for routers because it reduces overhead; the router doesn't have to store specific sender information, as every sender can also be a receiver, and vice versa. The local router acts as a Designated Forwarder (DF) to prevent loops by managing and organizing traffic flow within its segment. Examples of BIDIR-PIM scenarios include multiplayer online video games.

How are Rendezvous Points Configured in PIM Networks?

PIM networks have the following RP configuration options:

• Manual Selection: RPs can be manually configured. This option does not scale well and does not support redundant RP scenarios.
• BSR (Bootstrap Router): The Bootstrap Router, an IETF standard, facilitates RP (Rendezvous Point) discovery and distribution, not RP election itself. A BSR is elected from candidate BSRs based on configured priority or, if priorities are equal, the highest IP address. Candidate BSRs are configured to indicate their availability. Candidate RPs, which may or may not be the same routers as candidate BSRs, register with the BSR. The elected BSR sends messages to the all-PIM-routers group (224.0.0.13) to announce its presence and activity. These messages include the BSR's IP address, enabling candidate RPs to register via unicast. The BSR compiles a list of these candidate RPs and includes this list in its bootstrap messages. Routers, upon receiving these messages, use a hash algorithm, applied to this RP list, to select the best RP for each multicast group.
• Auto-RP: Auto-RP is Cisco's proprietary mechanism for dynamically discovering and distributing RP information. Mapping agents, configured on the network, play a key role in this process. While connected to multicast sources and responsible for registering them with the discovered RP, the mapping agent's primary function in Auto-RP is RP discovery and distribution. Candidate RPs signal their availability by joining the "RP-announce" group (224.0.1.39). Mapping agents also join this group to learn about the available RPs. The mapping agent collects these RP announcements and builds a mapping of multicast groups to candidate RPs. For each group, the mapping agent selects the most specific RP; if scopes are equally specific, the RP with the highest IP address is chosen. This group-to-RP mapping is then advertised by the mapping agent to the "RP-discovery" group (224.0.0.40), which Cisco routers automatically join to learn the designated RP for each multicast group. If a current RP fails, the process repeats, and the mapping agent selects the next most specific (or highest IP) candidate RP for that group.
• Anycast-RP: Anycast RP is a method for deploying multiple RPs to enhance redundancy and distribute the workload in a PIM-SM network. It leverages anycast addressing, where multiple RPs are each configured with a unique IP address, but they share a common anycast address. Routing protocols (like BGP or OSPF) are configured to recognize and advertise this anycast address. When a router needs to discover an RP, it sends a request to the anycast address. The routing protocol then directs this request to the nearest RP, based on routing metrics. This ensures that routers connect to the closest available RP. If an RP fails, the routing protocol automatically redirects requests to the next closest RP sharing the same anycast address. This provides redundancy and automatic failover. Because traffic is routed to the nearest RP, it helps distribute the load across the RP servers. To maintain consistent multicast information across all RPs, MSDP (Multicast Source Discovery Protocol) is typically used. MSDP allows the RPs to share information about active multicast sources, ensuring that each RP has a complete view of the multicast topology and preventing inconsistencies between different parts of the network.

What is Reverse Path Forwarding (RPF), and how does Multicast use it to Prevent loops?

RPF is a rule used by multicast to regulate the flow of multicast traffic. The rule states that if a router receives multicast traffic from a source, the router needs to verify that the interface it received the packet on would be the same interface it would use to send a unicast packet to the source, i.e., the receiving interface must be the same as the outgoing interface according to the IGP. If not, it drops the packet. This rule helps regulate traffic and prevent loops in the network by policing how traffic enters a segment.

Conclusion

In conclusion, IP multicast offers a powerful and efficient way to deliver data to multiple recipients simultaneously. By understanding the underlying mechanisms of group membership, distribution trees, and multicast routing protocols, network engineers can effectively leverage multicast to support a wide range of applications, from multimedia streaming to critical network infrastructure. As networks continue to evolve and the demand for bandwidth-efficient solutions grows, IP multicast will undoubtedly remain a cornerstone of modern networking.