2. Understanding Multicast Communication

Multicast at Scale

This chapter introduced the fundamental concepts of multicast communication and explained why multicast forwarding behaves differently from traditional Ethernet switching. Historically, many early media-over-IP deployments used Layer 2 multicast architectures based on VLANs and IGMP Snooping. While this approach can be effective in smaller environments, the behaviour of multicast at Layer 2 introduces several engineering considerations that become increasingly important as facilities scale.

Why IP Media Networks Use Multicast

There are three common methods of communication used in IP networks:

For streaming media applications, multicast offers significant advantages over unicast transport. If a video sender were required to transmit a separate stream to every receiver, bandwidth consumption would increase linearly with the number of destinations. For example, a single UHD ST 2110 video stream approaching 10 Gbit/s would require approximately 100 Gbit/s of transmission bandwidth if ten receivers needed the signal. Multicast eliminates this requirement. The sender transmits a single stream while the network performs replication only where necessary. This provides:

• Efficient bandwidth utilisation
• Reduced processing load on transmitters
• One-to-many signal distribution
• Flexible routing workflows
• Elimination of dedicated distribution amplifiers

For these reasons, multicast forms the foundation of technologies such as:

• AES67
• RAVENNA
• SMPTE ST 2110

Multicast as a Shared Network Resource

Unlike unicast traffic, multicast forwarding is not based upon a unique destination device. Instead, multicast traffic is delivered to a multicast group that receivers may subscribe to. This introduces a fundamental difference in how switches process multicast traffic.

Unicast Communication

With unicast traffic, there is a one-to-one mapping of Media Access Control (MAC) addresses to devices, whereby each network interface is assigned a globally unique MAC address.

• Every destination MAC address uniquely identifies a single device
• Switches learn devices via source MAC addresses and associate them with switch interfaces
• Traffic is forwarded only to the correct destination

Multicast Communication

To allow switches to distinguish multicast traffic from unicast and broadcast traffic, multicast packets use dedicated destination MAC and IP address ranges. Multicast MAC addresses are not used in the same way as with traditional unicast traffic. With multicast, MAC addresses are mapped to multicast IP addresses rather than the NICs that the frames will arrive on. A multicast MAC address is 48 bits long and is split into two 24-bit sections:

• Organizationally unique identifier (OUI)
• Multicast Address

The multicast address portion of the MAC address is derived from the lower 23 bits of the multicast IP address.

• A multicast MAC address represents a multicast group rather than an individual device
• Multiple receivers may subscribe to the same group
• Switches must maintain additional forwarding information to determine where multicast traffic should be delivered

Consider a sender transmitting to a multicast group:

At Layer 2, the multicast group could be represented as an IPv4 multicast MAC:

The switch therefore understands:

The issue is that multicast MAC addresses are never used as source MAC addresses. Switches learn MAC locations from source MAC addresses. Therefore, a switch cannot observe:

As a result, multicast becomes a shared network resource that requires additional control mechanisms.

Multicast Flooding

On unmanaged networks, or on managed networks where multicast control has not been enabled, multicast will be flooded to all ports in the same VLAN in which the frame was not received on. Network switches use a MAC address table to map different MAC addresses to different switch interfaces. Network switches build a MAC address table by examining the source MAC address of Ethernet frames as they arrive. This allows the switch to associate each learned MAC address with the interface on which it was observed.

When a switch receives a Layer 2 data frame, the following occurs:

1.  The switch inspects the source MAC address
   • If the MAC address is not already in its MAC table, it adds the MAC address along with the interface the frame was received on into the MAC table
2. The switch inspects the destination MAC address
   • If there is an entry for the MAC in the MAC table, it is forwarded to the associated interface
   • If there is not an entry, the frame is flooded to all interfaces on the VLAN that the frame was not received on

This works well for unicast traffic; however the same processes do not apply to multicast. Because a Multicast MAC address is only ever found in the destination field of a frame, the switch will never get the opportunity to map the multicast MAC address with a switchport (as per step one). This means that when a multicast frame arrives, the destination multicast MAC address will not exist in the switch’s MAC address table, and the frame will be flooded to all ports on a switch. 

Broadcast Domains and VLANs

Whenever a switch floods a frame, it only floods that frame within the local broadcast domain. A broadcast domain defines the collection of devices that may receive Layer 2 broadcast traffic. In Ethernet networks, broadcast domains are typically created using Virtual LANs (VLANs). A VLAN divides a physical switch network into multiple independent Layer 2 networks. Devices within the same VLAN can communicate directly at Layer 2, while devices in different VLANs require a router or Layer 3 switch to communicate.

Because unknown multicast traffic is also flooded within a VLAN, the size of the VLAN directly affects the number of devices that may receive unwanted multicast traffic. Smaller VLANs therefore reduce the scope of multicast flooding, while larger VLANs increase the potential impact. This relationship between VLAN size and multicast behaviour becomes increasingly important as media facilities scale. VLANs therefore provide the first level of multicast containment by limiting how far unknown multicast traffic can spread. However, they do not eliminate multicast flooding entirely. Additional multicast control mechanisms are required to determine which receivers actually require each multicast stream.

VLANs define the boundary over which Layer 2 traffic is distributed, but they also define the scope of many Layer 2 network events. As VLANs grow larger, the number of devices potentially affected by a configuration error, software defect, multicast flooding event or broadcast storm also increases. This area of potential impact is often referred to as the network’s failure domain or blast radius.

Multicast Addresses and Aliasing

One of the lesser-known characteristics of multicast is that Ethernet multicast addressing does not scale in the same way as IPv4 multicast addressing. Every multicast packet contains:

• A multicast Ethernet MAC address
• A multicast IP address

The multicast MAC address is derived from the lower 23 bits of the multicast IPv4 address. However, the multicast MAC address space is smaller than the IPv4 multicast address space. As a result:

This behaviour is known as multicast address aliasing. This occurs because the Ethernet multicast MAC address contains only 23 bits that uniquely identify a multicast group, whereas the IPv4 multicast address space contains 28 bits. As a result, five bits of addressing information are lost during the mapping process (2⁵ = 32). Because only the lower 23 bits of the IPv4 multicast address are encoded into the Ethernet MAC address, different multicast IP addresses inevitably share the same Ethernet destination MAC address.

The table above indicates a single multicast MAC address maps to 32 multicast IP addresses. In a network that relied solely on Layer 2 MAC forwarding, a receiver joining one multicast group could also receive traffic destined for the other 31 multicast IP addresses that share the same multicast MAC address. This happens because Layer 2 networks make forwarding decisions based only on the MAC addresses, so all 32 multicast IP addresses appear identical at Layer 2 and are forwarded to the same receiver. In practice, this means that Layer 2 forwarding decisions cannot always rely solely on Ethernet MAC addressing information. Modern enterprise and data centre switches typically mitigate this behaviour by inspecting Layer 3 multicast information when building multicast forwarding tables. However, address aliasing remains a fundamental characteristic of Ethernet multicast and becomes increasingly relevant as multicast scale increases. This is one of the reasons multicast forwarding requires additional logic beyond traditional MAC address learning. Modern switches therefore maintain dedicated multicast forwarding tables alongside their traditional MAC address tables, allowing forwarding decisions to consider both Layer 2 and Layer 3 multicast information.

This is one of the key reasons Layer 2 multicast forwarding cannot rely solely on traditional Ethernet switching behaviour. Additional multicast control protocols are required to determine which receivers actually require each multicast stream. Fortunately, modern switch platforms largely mitigate multicast address aliasing. However, the limitation highlights an important point: multicast forwarding is inherently more complex than unicast forwarding and requires dedicated control mechanisms beyond traditional Ethernet switching.

As facilities increase in size, network designers use several complementary techniques to control the scope of multicast traffic. VLANs reduce the size of Layer 2 broadcast domains, IGMP Snooping limits multicast forwarding within those domains, and routed multicast introduces Layer 3 boundaries that further reduce the impact of network events. These techniques are introduced progressively throughout the remainder of this guide.

In the next chapter, these multicast control mechanisms are introduced through IGMP, IGMP Snooping and IGMP Querier operation, showing how Layer 2 media networks learn which receivers require each multicast stream.

Continue to Chapter 3