Help us improve your experience.

Let us know what you think.

Do you have time for a two-minute survey?

Multicast Support in EVPN-VXLAN Overlay Networks

 

This topic describes the following multicast feature, which is supported in an EVPN-VXLAN overlay network:

IPv4 Inter-VLAN Multicast Forwarding Modes for EVPN-VXLAN Overlay Networks

Starting with Junos OS Release 17.3R3, the QFX10000 line of switches support two modes of IPv4 inter-VLAN multicast forwarding—centrally-routed mode and edge-routed mode—in an EVPN-VXLAN overlay network. The IP fabric architecture of your EVPN-VXLAN overlay network determines the mode that you must use.

For more information about these EVPN-VXLAN architectures, the multicast mode required for each architecture, and how the multicast forwarding modes work, see the following sections:

Benefits of IPv4 Inter-VLAN Multicast Forwarding Modes

  • A new configuration statement enables you to choose the inter-VLAN multicast forwarding mode that suits the architecture of your EVPN-VXLAN overlay network.

Supported EVPN-VXLAN Architectures and IPv4 Inter-VLAN Multicast Forwarding Modes

We support the following two modes of IPv4 inter-VLAN multicast forwarding:

  • The centrally-routed mode is supported only in EVPN-VXLAN centrally-routed bridging overlays (EVPN-VXLAN networks with a two-layer IP fabric). This overlay network is comprised of a layer of Layer 3 spine devices and another layer of Layer 2 leaf devices. All spine devices are configured with IRB interfaces that can route the multicast packets from one VLAN to another. However, only one spine device is elected to handle the inter-VLAN routing for the network.

  • The edge-routed mode is supported only in EVPN-VXLAN edge-routed bridging overlays (EVPN-VXLAN networks with a collapsed IP fabric). This overlay network is comprised of a layer of Layer 3 and 2 spine-leaf devices, all of which are configured with IRB interfaces. In this network, each of the spine-leaf devices handles the routing of multicast packets from one VLAN to another.

For centrally-routed bridging overlays, you can simply retain the default setting (local-remote) of the irb configuration statement at the [edit forwarding-options multicast-replication evpn] hierarchy level on all spine devices. For edge-routed bridging overlays, you must explicitly specify the local-only option for the irb configuration statement on all spine-leaf devices.

Note

We do not recommend specifying the local-remote option on some QFX10000 switches and the local-only option on the other QFX10000 switches in either of the overlay networks. Doing so might cause the QFX10000 switches to forward the inter-VLAN multicast traffic inconsistently.

For the edge-routed bridging overlay, we support both Junos Fusion Data Center and non-Junos Fusion Data Center deployments. Although the Junos Fusion Data Center with EVPN architecture actually includes a layer of aggregation devices and a layer of satellite devices, only the aggregation devices actually run EVPN. As a result, from the perspective of inter-VLAN multicast forwarding, the Junos Fusion Data Center with EVPN architecture is considered to be an edge-routed bridging overlay.

Understanding Multicast Traffic Flows in a Centrally-Routed Bridging Overlay

This section describes the multicast traffic flows in a centrally-routed bridging overlay.

The network shown in Figure 1 includes the following devices.

Figure 1: Centrally-Routed Bridging Overlay
 Centrally-Routed Bridging Overlay
  • Two QFX10000 switches that function as Layer 3 spine devices, on which the following key features are configured:

    • Centrally-routed (local-remote) multicast forwarding mode.

    • Protocol Independent Multicast (PIM). By virtue of PIM hello messages, Spine 1 is elected as the PIM designated router (PIM DR).

    • VLANs A and B.

      Note

      For inter-VLAN multicast forwarding to work properly in this scenario, you must configure all VLANs on each spine device.

    • IRB interfaces associated with VLANs A and B.

  • Four QFX5100 switches that function as Layer 2 leaf devices, on which VLANs A and B are configured are follows:

    • Leafs 1 and 3 are configured with VLAN A only.

    • Leaf 2 is configured with VLAN B only.

    • Leaf 4 is configured with VLANs A and B.

  • A multicast source and various receivers.

When the multicast source shown in Figure 1 sends a packet from VLAN A, the following flows occur:

  • Flow 1: Intra-VLAN traffic—Based on the ingress replication mechanism, Leaf 1 replicates and switches the packet to all spine and other leaf devices. Leafs 3 and 4, on which VLAN A is configured, receive and forward the packet to the connected multicast receivers.

  • Flow 2: Inter-VLAN traffic—Upon receipt of the packet from Leaf 1 as described in flow 1, Spine 1, which is the PIM DR, takes the following action:

    • Routes the packet over the IRB interface associated with VLAN B.

    • Based on the ingress replication mechanism, replicates and forwards the packet to the other spine and the leaf devices.

      Leafs 2 and 4, on which VLAN B is configured, receive and forward the packet to the connected multicast receivers.

Understanding Multicast Traffic Flows in an Edge-Routed Bridging Overlay

This section describes the multicast traffic flow in an edge-routed bridging overlay.

The non-Junos Fusion Data Center network shown in Figure 2 includes the following devices.

Note

The information in this section also applies to a Junos Fusion Data Center with EVPN network except that instead of the QFX10002 switches functioning as spine-leaf devices, they function as aggregation devices. See Multicast Forwarding at Layer 2 in a Junos Fusion Data Center with EVPN for details on multicast forwarding and replication in a Junos Fusion Data Center network.

Figure 2: Edge-Routed Bridging Overlay (Non-Junos Fusion Data Center)
Edge-Routed Bridging Overlay
(Non-Junos Fusion Data Center)
  • Four QFX10002 switches that function as Layer 3 and Layer 2 spine-leaf devices, on which the following key features are configured:

    • Edge-routed (local-only) multicast forwarding mode.

    • PIM. To support the edge-routed mode of multicast forwarding, each spine-leaf device must act as the PIM DR for each VLAN. To enable a spine-leaf device to elect itself as the PIM DR, for each IRB interface, specify the distributed-dr configuration statement at the [edit protocols pim interface interface-name] hierarchy.

    • VLANs A and B.

      Note

      For inter-VLAN multicast forwarding to work properly in this scenario, you must configure all VLANs on each spine-leaf device.

    • IRB interfaces associated with VLANs A and B.

  • A multicast source and various receivers.

When the multicast source shown in Figure 2 sends a packet from VLAN A, the following flows occur:

  • Flow 1: Intra-VLAN traffic—Based on the ingress replication mechanism, Spine-Leaf 1 replicates and switches the packet to the other spine-leaf devices. The spine-leaf devices forward the packet to VLAN A. Further, Spine-Leafs 2 and 3 forward the packet to VLAN A receivers.

  • Flow 2: Inter-VLAN traffic—Upon receipt of the packet from Spine-Leaf 1 as described in flow 1, the spine-leaf devices route the packet over the IRB interface associated with VLAN B. Further, Spine-Leafs 2 and 4 forward the packet to the VLAN B receivers.

Differences Between IPv4 Inter-VLAN Multicast Forwarding Modes

There is an important difference between the centrally-routed and edge-routed modes.

With centrally-routed mode, for each VLAN, the spine device elected as the PIM-DR must replicate and send the packet to the other devices in the network. Keep in mind that the additional instances of replication consume bandwidth, and if many VLANs are configured, can potentially flood the EVPN core with multicast packets.

With edge-routed mode, the first spine-leaf device to receive a multicast packet replicates and sends the packet to the other spine-leaf devices. Upon receipt of the packet, the other spine-leaf devices route the packet to each VLAN and send the packet to access-side interfaces that handle traffic for the multicast receivers. In other words, the spine-leaf devices do not replicate the packet and send the packet out of the EVPN core-facing interfaces, which prevents excessive bandwidth consumption and congestion in the EVPN core.

Release History Table
Release
Description
Starting with Junos OS Release 17.3R3, the QFX10000 line of switches support two modes of IPv4 inter-VLAN multicast forwarding—centrally-routed mode and edge-routed mode—in an EVPN-VXLAN overlay network. The IP fabric architecture of your EVPN-VXLAN overlay network determines the mode that you must use.