Optimized Intersubnet Multicast in EVPN Networks

Benefits of OISM

• Enables EVPN-VXLAN fabrics with the ERB overlay model to support multicast traffic with sources and receivers outside of the fabric.
• Minimizes multicast control packets and replicated data packets in the EVPN fabric core to optimize fabric multicast performance in scaled designs.

OISM Support in EVPN Instances

We support OISM in EVPN-VXLAN fabrics in the following types of EVPN instances:

• EVPN in the default switch instance:

  • Starting in Junos OS Release 21.2R1 on QFX5110, QFX5120, and QFX10002 (except QFX10002-60C) switches.

  • Starting in Junos OS Release 22.2R1 on EX4650, QFX10008, and QFX10016 switches.

  • Starting in Junos OS Release 22.3R1 on EX4300-48MP and EX4400 switches.

  • Starting in Junos OS Release 23.4R1 on ACX7024, ACX7100-32C, and ACX7100-48L routers.

• MAC-VRF EVPN routing instances with vlan-aware and vlan-based service types only (see MAC-VRF Routing Instance Type Overview):

  • Starting in Junos OS Evolved Release 22.1R1 on QFX5130-32CD and QFX5700 switches.

  • Starting in Junos OS Release 22.2R1 on EX4650, QFX5110, QFX5120, QFX10002 (except QFX10002-60C), QFX10008, and QFX10016 switches.

  • Starting in Junos OS Evolved Release 22.3R1 on PTX10001-36MR, PTX10004, PTX10008, and PTX10016 routers.

  • Starting in Junos OS Release 23.4R1 on ACX7024, ACX7100-32C, and ACX7100-48L routers.

On Junos OS Evolved devices, we support EVPN-VXLAN using EVPN configurations with MAC-VRF instances only, and not in the default switch instance. As a result, on these devices we support OISM only in MAC-VRF EVPN instances.

OISM with Multicast Protocols and Other Multicast Optimizations in EVPN Fabrics

OISM works with the following multicast protocols and other EVPN multicast optimization features:

• Multicast Protocols Supported with OISM
• Other Multicast Optimization Features That Work with OISM

OISM Device Roles

Figure 1 shows a simple EVPN-VXLAN ERB overlay fabric and the OISM device roles in the fabric.

Figure 1: EVPN Fabric with OISM

Table 1 summarizes the device roles.

See Configuration Elements for OISM Devices (Symmetric Bridge Domains) for details on the configuration elements that are common and those that are different for each device role.

PIM Domain with External Multicast Sources and Receivers

In Figure 1, the OISM border leaf devices connect to multicast sources and receivers outside the EVPN fabric in a representative external PIM domain. The multicast devices in the external PIM domain follow standard PIM protocol procedures; their operation is not specific to OISM. External multicast traffic flows at L3 through the PIM domain.

You can use OISM to route and to forward multicast traffic in an EVPN-VXLAN ERB overlay fabric between devices in the following use cases:

• Internal multicast sources and receivers

• Internal multicast sources and external multicast receivers

• External multicast sources and internal multicast receivers

For simplicity, in this documentation we represent the external PIM domain as:

• A PIM router (a device such as an MX Series router) that doubles as the PIM rendezvous point (RP).

• An external source.

• An external receiver.

Supported Methods for Multicast Data Transfer to or from an External PIM Domain

OISM border leaf devices support one or more methods to route multicast traffic to and from devices outside of the fabric. Supported methods are platform-dependent.

Some platforms don't support the border leaf role. If you don't see a platform listed in Table 2 in the Supported Platforms column for any of the external multicast methods, that means the platform doesn't support the border leaf role.

See Overview of Multicast Forwarding with IGMP Snooping or MLD Snooping in an EVPN-VXLAN Environment for a general overview of connecting EVPN-VXLAN fabrics to an external PIM domain using a L2 M-VLAN or L3 links.

OISM Bridge Domains (VLANs)

Table 3 summarizes the OISM bridge domains or VLANs and describes how OISM uses them.

The OISM implementation uses a symmetric bridge domains model. The following figure shows how you set up the bridge domains and VLANs in the EVPN fabric on the OISM devices with this model.

Figure 2: EVPN Fabric with OISM Symmetric Bridge Domains Model

In the symmetric bridge domains model, you must configure the SBD and all revenue bridge domains on all OISM border leaf and server leaf devices.

You also configure the following VLANs uniformly on the border leaf devices that connect to the external PIM domain:

• The M-VLAN, if you use the M-VLAN IRB method for external multicast.

• A unique extra non-EVPN VLAN on each border leaf device, if you use the non-EVPN IRB method for external multicast.

The lean spine devices in the fabric usually serve only as IP transit devices and possibly as route reflectors. As a result, you don't usually need to configure these elements on the lean spine devices. (See the Lean Spine row in Table 1 for some exceptions.)

Configuration Elements for OISM Devices (Symmetric Bridge Domains)

This section summarizes the elements you need to configure on:

• All OISM devices—Devices in the border leaf role and the server leaf role, and spine devices that also serve as AR replicators in standalone mode when you integrate AR with OISM.

  See Table 4.

• Server leaf devices only.

  See Table 5.

• Border leaf devices only, based on the method you use for external multicast:

  • M-VLAN IRB interface

  • Classic L3 interface

  • Non-EVPN IRB interface

  See Table 6.

Some elements are optional, which the description notes.

Table 4 lists the elements you configure on all OISM devices.

Table 5 lists the elements you configure on the server leaf devices.

Table 6 lists the elements you configure on the border leaf devices based on the external multicast method you use.

See the following sections for more details on configuring OISM devices:

• Configure Common OISM Elements on Border Leaf Devices and Server Leaf Devices (Symmetric Bridge Domains)

• Configure Server Leaf Device OISM Elements (Symmetric Bridge Domains)

• Configure Border Leaf Device OISM Elements with M-VLAN IRB Method (Symmetric Bridge Domains)

• Configure Border Leaf Device OISM Elements with Classic L3 Interface Method (Symmetric Bridge Domains)

• Configure Border Leaf Device OISM Elements with Non-EVPN IRB Method (Symmetric Bridge Domains)

For a full OISM configuration example of a data center fabric use case that includes classic L3 interface connections to the external PIM domain, see Optimized Intersubnet Multicast (OISM) with Assisted Replication (AR) for Edge-Routed Bridging Overlays.

Local Routing on OISM Devices

In Figure 3, we illustrate how local routing and forwarding works in general on OISM devices. As the figure shows, OISM local routing forwards the traffic on the source VLAN. Each leaf device routes the traffic locally to its receivers on other VLANS, which avoids hairpinning for intersubnet routing on the same device.

Figure 3: Local Routing with OISM

In this case, the source traffic comes from Mcast-Src-1 on VLAN-1, the blue VLAN. Server leaf devices use IRB interfaces and PIM in passive mode to route traffic between VLANs. With PIM in passive mode, server leaf devices:

• Don’t become PIM neighbors with the other leaf devices.

• Act as a local PIM RP, create local PIM state upon receiving IGMP or MLD reports, and avoid doing source registration.

As a result, the server leaf devices forward and route multicast traffic within the fabric as follows to receivers interested in the multicast group:

• The ingress leaf device (Leaf-1) forwards the traffic on the source VLAN into the EVPN fabric toward the other leaf devices with interested receivers.

• All of the server leaf devices don't need to forward the traffic back into the EVPN core to another device that is a designated router. Server leaf devices can locally:

  • Forward the traffic on the source VLAN toward local interested receivers on the source VLAN.

  • Route the traffic from the source VLAN through the IRB interfaces toward local interested receivers in other VLANs.

Multicast Traffic Forwarding and Routing with Source and Receivers Inside the EVPN Data Center

When the multicast source is inside the EVPN fabric, the server leaf devices receive the multicast traffic on the source VLAN. Then they locally route or forward the traffic as described in Local Routing on OISM Devices.

The following figure illustrates OISM local routing and forwarding within an EVPN fabric in detail. The figure also shows how local routing works with EVPN multihoming for a multicast receiver.

Figure 4: OISM with an Internal Multicast Source and Internal Multicast Receivers

In Figure 4, the multicast source, Mcast-Src-1, is single-homed to Leaf-1. The source VLAN is VLAN-1 (the blue VLAN). Multicast control and data traffic flow proceeds as follows:

1. Receivers on all three server leaf devices sent IGMP or MLD reports (join messages) expressing interest in receiving the traffic for a multicast group.

2. Leaf-1 forwards the traffic on the source VLAN to both Leaf-2 and Leaf-3 because both leaf devices have interested receivers. In this case, the receivers on Leaf-2 and Leaf-3 use single-homing.

3. Leaf-2 and Leaf-3 forward or locally route the traffic to their interested receivers (Rcvr-2, Rcvr-3, and Rcvr-4) as described in Local Routing on OISM Devices.

4. Rcvr-1 on VLAN-2 is multihomed to Leaf-1 and Leaf-2 in an EVPN ES. Rcvr-1 has expressed interest in receiving the multicast traffic, so:

   • Both server leaf devices, Leaf-1 and Leaf-2, receive the IGMP or MLD report.
   • Both Leaf-1 and Leaf-2 locally route the traffic from the source VLAN (VLAN-1) because each device has the PIM passive mode configuration.
   • However, because Leaf-1 is the DF for the EVPN ES, only Leaf-1 forwards the traffic to Rcvr-1.

5. The border leaf devices receive the multicast traffic through the EVPN fabric on the source VLAN. Note that the border leaf devices could have local receivers, although we don't show that case. With local receivers, the device also locally routes or forwards the traffic to those receivers the same way the server leaf devices do.

Figure 4 also shows that the border leaf devices locally route the traffic from the source VLAN toward any external multicast receivers in the external PIM domain. See Supported Methods for Multicast Data Transfer to or from an External PIM Domain for the available external multicast methods by platform. Later sections describe the multicast control and data traffic flow for external source and external receiver use cases.

Multicast Traffic From an Internal Source to Receivers Outside the EVPN Data Center—M-VLAN IRB Method

In Figure 5, we illustrate the OISM use case where a multicast source inside the EVPN fabric sends multicast traffic to an interested receiver outside the fabric using the M-VLAN IRB method for external multicast. (Supported Methods for Multicast Data Transfer to or from an External PIM Domain lists external multicast method support by platform.)

The border leaf devices you configure in the OISM PEG role receive the multicast traffic on the source VLAN through the EVPN core. Then the border leaf devices replicate the traffic and route it onto the M-VLAN toward the external PIM domain to reach the external receiver.

Figure 5: OISM with an Internal Multicast Source and an External Multicast Receiver—M-VLAN IRB Method

In Figure 5, the internal source for a multicast group is Mcast-Src-2, the same device as Rcvr-1, which is multihomed to Leaf-1 and Leaf 2. The source sends the multicast traffic on VLAN-2. The external receiver, Ext-Mcast-Rcvr, expresses interest in receiving the multicast traffic for that multicast group (sends a join message). Internal receivers Rcvr-3 (on VLAN-1) and Rcvr-4 (on VLAN-2) also request to join the multicast group and receive the traffic.

Note that the PIM router is multihomed to both BL-1 and BL-2, the PEG devices, in the EVPN fabric. Those connections are in the same ES; the DF election process chooses one of these devices as the DF for the ES. Only the DF will forward traffic (on the M-VLAN) toward external receivers.

The source traffic reaches the interested internal and external receivers as follows:

• Traffic Flow from Multihomed Source to Internal Receivers
• Traffic Flow to External Receiver—M-VLAN IRB Method

Multicast Traffic from an Internal Source to Receivers Outside the EVPN Data Center—L3 Interface Method or Non-EVPN IRB Method

In Figure 6, we illustrate the OISM use case where a multicast source inside the EVPN fabric sends multicast traffic to a receiver outside the fabric using either of the following methods for external multicast:

• Classic L3 interface external multicast method:

  On each border leaf device, you configure a classic L3 interface with family inet that connects to the external PIM router. You assign an IP address to that interface on a different subnet than the L3 interface subnets on other border leaf devices in the fabric.

  You enable PIM on the interface and include the interface in the tenant VRF instances that have multicast data receivers. This method differs from the M-VLAN IRB method because you don't extend this interface in the EVPN instance.

  Note:

  The L3 interface connection here can be an individual physical interface or an AE interface bundle that includes multiple physical L3 interfaces.

• Non-EVPN IRB external multicast method:

  On each border leaf device, you configure a unique extra VLAN that is only for external multicast. You also configure a corresponding L3 IRB interface with an IP address that connects to the external PIM router. The extra VLAN ID can't be the same as the VLAN ID of the revenue bridge domains, SBD, or extra VLAN on any other border leaf device in the fabric. In addition, similar to the L3 interface method, the non-EVPN IRB interfaces on different border leaf devices should connect to the PIM router on different subnets in the fabric.

  You enable PIM on the IRB interface and include the interface in the tenant VRF instances that have multicast data receivers. This method differs from the M-VLAN IRB method because you don't extend this VLAN or IRB interface in the EVPN instance.

Supported Methods for Multicast Data Transfer to or from an External PIM Domain lists the platforms that support these external multicast methods.

Figure 6 includes the same internal multihomed source, internal receivers, and an external receiver as in Multicast Traffic From an Internal Source to Receivers Outside the EVPN Data Center—M-VLAN IRB Method. See the steps in Traffic Flow from Multihomed Source to Internal Receivers for details on the internal multicast traffic flow. In this section we describe only what's different in this case, which is the multicast traffic flow from the border leaf devices to the external receiver.

Figure 6: OISM with an Internal Multicast Source and an External Multicast Receiver—L3 Interface or Non-EVPN IRB Method

In Figure 6, the internal source for a multicast group is Mcast-Src-2, which is multihomed to Leaf-1 and Leaf-2. The source sends the multicast traffic on VLAN-2, the red VLAN. The external receiver, Ext-Mcast-Rcvr, expresses interest in receiving the multicast traffic for that multicast group (sends a join message).

The external multicast flow in this use case is very similar to the M-VLAN IRB use case. The border leaf devices in the OISM PEG role receive the multicast traffic on the source VLAN through the EVPN core. However, the main difference in this case is that the external multicast interfaces don't use EVPN signaling and don't share an ESI across the border leaf devices. The external multicast interfaces on each border leaf device are distinct, and each have L3 reachability to the external PIM gateway router. The border leaf device that establishes the PIM join state replicates and sends the traffic on the L3 interface or non-EVPN IRB interface to the external PIM domain with the external receiver.

The following section explains how the source traffic reaches the interested external receiver.

Multicast Traffic from an External Source to Receivers Inside the EVPN Data Center—M-VLAN IRB Method

Figure 7 illustrates the OISM use case where a multicast source outside the EVPN fabric sends multicast traffic to receivers inside the fabric. This use case exhibits the two main ways OISM uses the SBD in the EVPN core:

• To carry external multicast source traffic.

• To advertise SMET Type 6 routes.

The Type 6 routes ensure that the border leaf devices only forward the traffic toward the EVPN devices with interested receivers.

OISM border leaf devices receive external multicast source traffic through the M-VLAN IRB interfaces. OISM devices use the SBD to forward traffic toward EVPN server leaf devices with interested receivers on the revenue bridge domains. Each leaf device then locally forwards or routes the traffic on the revenue bridge domains to its local receivers.

This use case has an internal receiver that is multihomed to two server leaf devices.

Figure 7: OISM with an External Multicast Source and an Internal Multihomed Multicast Receiver—M-VLAN IRB Method

In Figure 7, Rcvr-1 within the EVPN fabric is multihomed to server leaf devices Leaf-1 and Leaf-2. Rcvr-1 expresses interest in receiving traffic from a multicast group. The source for the multicast traffic for the group is Ext-Mcast-Src in the external PIM domain.

The external source traffic reaches the interested multihomed receiver, Rcvr-1, as follows:

• Multicast Control Flow between the Internal Multihomed Receiver and the External Source—M-VLAN IRB Method
• Traffic Flow from the Border Leaf Devices to Internal Receivers—M-VLAN IRB Method
• What Happens When a Multihomed External PIM Router Load-Balances the Traffic—M-VLAN IRB Method
• What Happens with Local Receivers on Border Leaf Devices—M-VLAN IRB Method

Multicast Traffic from an External Source to Receivers Inside the EVPN Data Center—L3 Interface Method or Non-EVPN IRB Method

Figure 8 illustrates the OISM use case where a multicast source outside the EVPN fabric sends multicast traffic to receivers inside the fabric. In this case, the fabric uses the classic L3 interface or non-EVPN IRB external multicast method to connect to the external PIM domain. Also, this case includes the following internal receivers:

• A receiver that is multihomed to two server leaf devices.

• A local receiver on one of the border leaf devices.

This use case, like the M-VLAN IRB external source use case in Figure 7, shows the two main ways OISM uses the SBD in the EVPN core—To carry external multicast source traffic and to advertise SMET Type 6 routes. The Type 6 routes ensure that the border leaf devices only forward the traffic toward the EVPN devices with interested receivers.

Figure 8: OISM with an External Multicast Source and an Internal Multihomed Multicast Receiver—L3 Interface or Non-EVPN IRB Method

In Figure 8:

• Rcvr-1 is multihomed to server leaf devices Leaf-1 and Leaf-2 in the EVPN fabric, and expresses interest in receiving traffic from a multicast group.

• Rcvr-5 on BL-2 in the EVPN fabric is also interested in receiving the multicast traffic.

• Ext-Mcast-Src in the external PIM domain is the source for the traffic for the multicast group.

The multicast control flow and data traffic flow for external multicast are similar for the classic L3 interface and the non-EVPN IRB interface methods. As a result, in this section we commonly say external multicast interface when we refer to the border leaf device external connection points.

The external source traffic reaches the interested receivers (Rcvr-1 and Rcvr-5) as follows:

• Multicast Control Flow between the Internal Receivers and the External Source—L3 Interface or Non-EVPN IRB Method
• Traffic Flow from the Border Leaf Devices to Internal Receivers—L3 Interface or Non-EVPN IRB Method

AR and OISM with an Internal Multicast Source

In Figure 9, we show an OISM use case where you configure the spine devices as standalone AR replicator devices. The OISM server leaf and border leaf devices are AR leaf devices. The AR replicator devices handle replicating the multicast traffic for the OISM server leaf and border leaf devices. This case shows a multicast source and single-homed receivers behind server leaf devices inside the EVPN fabric.

With OISM traffic from an internal source, the ingress device forwards the traffic in the EVPN fabric on the source VLAN. When you also enable AR, the ingress leaf device forwards one copy of the traffic to an AR replicator. The AR replicator replicates the traffic and sends the copies on the source VLAN to the other leaf devices with interested receivers. Then each leaf device:

• Locally forwards the traffic toward its receivers on the source VLAN.

• Locally routes the traffic toward its receivers on the other revenue VLANs.

Figure 9: AR with OISM—Internal Multicast Source and Single-Homed Internal Receivers

In the use case in Figure 9:

1. Rcvr-2, Rcvr-3 and Rcvr-4 send IGMP or MLD reports to join the multicast group.

2. The traffic source for the multicast group, Mcast-Src-1, forwards the traffic on the source VLAN, VLAN-1, to Leaf-1.

3. Leaf-1 forwards the traffic to one of the available AR replicators on VLAN-1 to replicate the traffic to the other leaf devices with interested receivers. In this case, Leaf-1 forwards the traffic to ARR-1.

   Note:

   See AR Leaf Device Load Balancing with Multiple Replicators for details on how AR leaf devices load-balance among multiple available AR replicators.

4. ARR-1 replicates the traffic and sends copies on VLAN-1 toward all the leaf devices with interested receivers.

5. Each server leaf device:

   • Forwards the traffic toward the interested receivers on VLAN-1.

   • Locally routes the traffic to VLAN-2 and forwards it toward the interested receivers on VLAN-2.

   Also note that in Figure 9, Rcvr-1 is multihomed to Leaf-1 and Leaf-2, with Leaf-1 as the ESI DF. As a result, only Leaf-1 forwards the traffic to Rcvr-1 on VLAN-2.

6. If any external receivers expressed interest in receiving the traffic, the border leaf devices locally route the traffic to the external multicast interface. The external multicast interface sends the traffic toward any interested external receivers based on the external multicast method you configure.

See Configure Assisted Replication for details on how to configure AR.

AR and OISM with an Internal Multicast Source and Multihomed Receiver

In Figure 10, we show an OISM use case similar to the setup in AR and OISM with an Internal Multicast Source. However, in this case, the multicast source is behind a server leaf device that also has a multihomed receiver. In this case, AR operates in extended AR mode by default to efficiently support the multihomed receiver. See Extended AR Mode for Multihomed Ethernet Segments for full details on this mode.

Here is a summary of how multicast traffic ingressing on a server leaf device reaches a multihomed receiver in this case:

• The ingress server leaf device with an ESI for a multihomed receiver maintains a list of its multihoming peer leaf devices on the ES.

  The AR replicator device also knows which AR leaf devices have multihoming peers.

• The ingress server leaf device takes care of replicating and forwarding the multicast traffic to any of its multihoming peers that are interested in the traffic.

  The ingress leaf device also sends one copy to an AR replicator device to handle replication and forwarding to any other leaf devices.

Other than this difference in handling replication to the multihoming peers, the traffic flow to an AR replicator and then to the interested receivers is the same as we describe in AR and OISM with an Internal Multicast Source.

Figure 10: AR with OISM—Internal Multicast Source and Multihomed Receiver

In Figure 10:

1. Rcvr-1, Rcvr-2, Rcvr-3 and Rcvr-4 send IGMP or MLD reports to join the multicast group.

2. The traffic source for the multicast group, Mcast-Src-1, forwards the traffic on the source VLAN, VLAN-1, to Leaf-1.

3. Following the extended AR mode for multihoming peers, Leaf-1 forwards the traffic directly to its multihoming peer Leaf-2, which has an interested receiver. Leaf-1 uses the source VLAN, VLAN-1, according to OISM behavior.

4. Leaf-1 also forwards the traffic to one of the available AR replicators, ARR-1 in this case, on the source VLAN, VLAN-1.

   Note:

   See AR Leaf Device Load Balancing with Multiple Replicators for details on how AR leaf devices load-balance among multiple available AR replicators.

5. ARR-1 replicates the traffic and sends copies on VLAN-1 only toward the other leaf devices with interested receivers besides Leaf-2. Due to the default extended AR mode behavior (see Step 3 above), ARR-1 skips sending the traffic to Leaf-2, the multihoming peer of the ingress leaf device Leaf-1.

6. Each server leaf device then forwards or routes the traffic to its interested receivers.

   Note that in Figure 10, Leaf-1 is the ESI DF for multihomed receiver Rcvr-2. As a result, only Leaf-1 forwards the traffic to Rcvr-1 on VLAN-2.

See Configure Assisted Replication for details on how to configure AR.

AR and OISM with an External Multicast Source

In Figure 11, we show an OISM use case where you configure the spine devices as standalone AR replicator devices. The OISM server leaf and border leaf devices are AR leaf devices. The multicast source is outside the EVPN fabric in an external PIM domain. The border leaf devices in this case use the classic L3 interface method to connect to the PIM router and PIM RP.

With OISM traffic from an external source, the ingress border leaf device forwards the traffic in the EVPN fabric on the SBD VLAN. When you also enable AR, the ingress border leaf device forwards one copy of the traffic to an AR replicator. The AR replicator replicates the traffic and sends the copies on the SBD VLAN to the other leaf devices with interested receivers. Then each device locally routes the traffic it receives on the SBD toward its receivers on the revenue bridge domain VLANs.

Figure 11: AR with OISM-External Multicast Source

In the use case in Figure 11:

1. Rcvr-1 (multihomed to Leaf-1 and Leaf-2) and Rcvr-5 (a local host behind BL-2) send IGMP or MLD reports to join the multicast group.

2. The external source, Ext-Mcast-Src, sends multicast traffic through the external PIM domain. In this case we use the classic L3 interface external multicast method, and both devices sent a PIM join message, so the PIM router sends the traffic to both BL-1 and BL-2. (See Multicast Traffic from an External Source to Receivers Inside the EVPN Data Center—L3 Interface Method or Non-EVPN IRB Method for a full explanation of this behavior.)

   Note:

   As Figure 11 shows, in this use case, because BL-2 has a local receiver, BL-2 routes the incoming externally sourced traffic directly toward its receiver on VLAN-2. BL-2 doesn't route traffic to the local receiver that it receives on the SBD from ARR-2 because its reverse-path forwarding toward the external source refers to the L3 interface.

   BL-2 also doesn't route the traffic to the SBD because BL-1 is the PIM DR on the SBD (see the next step).

3. BL-1 is the PIM DR for the SBD, so BL-1 is the border leaf device that routes the externally sourced traffic into the EVPN fabric. With AR enabled, BL-1 forwards the traffic on the SBD to one of the available AR replicators. In this case, BL-1 forwards the traffic to ARR-2.

   Note:

   See AR Leaf Device Load Balancing with Multiple Replicators for details on how AR leaf devices load-balance among multiple AR replicators.

4. ARR-2 replicates the traffic and sends copies on the SBD toward the leaf devices with interested receivers—in this case, Leaf-1, Leaf-2, and BL-2.

5. Each leaf device that receives the traffic on the SBD locally routes the traffic toward the interested receivers on the revenue VLANs. In this case:

   • BL-2 routes the traffic toward its receiver on VLAN-2.

   • Both Leaf 1 and Leaf-2 receive the traffic on the SBD. Rcvr-1 is multihomed to Leaf-1 and Leaf-2, and Leaf-1 is the ESI DF. As a result, only Leaf-1 forwards the traffic toward Rcvr-1 on VLAN-2.

See Configure Assisted Replication for details on how to configure AR.

IGMPv2 and IGMPv3 (or MLDv1 and MLDv2) in the Same EVPN-VXLAN Fabric

You have several options to configure IGMP snooping with IGMPv2, IGMPv3, or both IGMP versions together in an EVPN-VXLAN fabric with OISM. The same is true of MLD snooping with MLDv1, MLDv2, or both MLD versions together. You might also want to mix configuring IGMP and MLD together in the same fabric. This section includes configuration considerations for a few of these options.

If you have traffic for either IGMPv2 or IGMPv3 on a device with OISM enabled, you can enable that IGMP version globally on the device with IGMP snooping. You can alternatively enable that IGMP version only for the interfaces that will handle the multicast traffic. You can enable IGMP snooping with that version of IGMP on all VLANs or on specific VLANs, as needed.

You have the same options for either MDLv1 or MLDv2 with MLD snooping (on the platforms that support MLD with OISM).

You can also enable one version of IGMP with IGMP snooping and one version of MLD with MLD snooping together on the device with OISM.

However, OISM supports IGMP snooping with both IGMPv2 and IGMPv3 traffic together on a device only within the following constraints:

• You can't enable IGMP snooping with both IGMPv2 and IGMPv3 for interfaces in the same VLAN.

• You can't enable IGMP snooping with both IGMPv2 and IGMPv3 for VLANs that are part of the same L3 VRF instance with OISM enabled.

The constraints above also apply if you want to enable MLD snooping with both MLDv1 and MLDv2 traffic together on a device.

These constraints don't apply if you use one version of IGMP with one version of MLD together on a device.

To support IGMP snooping with both IGMP versions, or MLD snooping with both MLD versions, you must configure:

• One tenant VRF instance to support the IGMPv2 or MLDv1 receivers.

• Another tenant VRF instance to support the IGMPv3 or MLDv2 receivers.

as follows:

1. In your configuration, define VLANs for the IGMPv2 receivers, and define different VLANs for the IGMPv3 receivers.

   Similarly, for MLD, define VLANs for the MLDv1 receivers, and define different VLANs for the MLDv2 receivers.

2. Include the IRB interfaces that support IGMPv2 in one VRF instance, and enable IGMPv2 on those IRB interfaces. Enable IGMP snooping on the corresponding VLANs.

   Similarly, for MLD, include the IRB interfaces that support MLDv1 in one VRF instance, and enable MLDv1 on those IRB interfaces. Enable MLD snooping on the corresponding VLANs.

3. Include the IRB interfaces that support IGMPv3 in another VRF instance, and enable IGMPv3 on those IRB interfaces. Enable IGMP snooping with the evpn-ssm-reports-only option on the corresponding VLANs.

   Similarly, for MLD, include the IRB interfaces that support MLDv2 in another VRF instance, and enable MLDv2 on those IRB interfaces. Enable MLD snooping with the evpn-ssm-reports-only option on the corresponding VLANs.

In this use case, for each IGMP or MLD version, allocate a set of VLANs and IRB interfaces for:

• The OISM revenue bridge domains.

• The SBD.

• Any external multicast VLANs and interfaces (depending on the external multicast method you use).

You also define two L3 VRF instances for each tenant instance you need in your installation, one for each IGMP version or MLD version. If you use MAC-VRF routing instances at L2, you might want to allocate different MAC-VRF EVPN instances for the IGMP snooping or MLD snooping traffic for each IGMP or MLD version.

The next sections show example configurations with both versions of IGMP or both versions of MLD together. You can scale these simple scenarios to support different tenants with different combinations of IGMP or MLD versions.

See Supported IGMP or MLD Versions and Group Membership Report Modes for more information on IGMP any-source multicast (ASM) mode and source-specific multicast (SSM) mode support with IGMPv2, IGMPv3, MLDv1, and MLDv2 in EVPN-VXLAN fabrics.

• Example Configuration with IGMPv2 and IGMPv3 Together
• Example Configuration with MLDv1 and MLDv2 Together

Latency and Scaling Trade-Offs for Installing Multicast Routes with OISM (install-star-g-routes Option)

Devices in an OISM-enabled fabric send EVPN Type 6 routes so other EVPN devices learn about receivers that are interested in the traffic for a multicast group. The receivers are in different OISM revenue bridge domains in an L3 VRF instance. To save bandwidth in the EVPN fabric core, OISM devices send and receive the Type 6 routes only on the OISM SBD in the routing instance.

To help minimizes packet loss at the onset of a multicast flow, we provide the install-star-g-routes option at the [edit <routing-instances name> multicast-snooping-options oism] hierarchy level (see oism (Multicast Snooping Options)). When you configure this option, upon receiving a Type 6 route, the RE on the device immediately installs corresponding (*,G) multicast routes on the PFE for all of the revenue bridge domain VLANs in the routing instance.

With this option, you trade off taking up extra PFE resources to improve network latency. Lower-scale deployments might have fewer multicast flows but have strict network latency requirements. To improve network latency in that case, the device installs the (*,G) routes in the data plane in advance of any incoming multicast traffic.

Configure this option:

• Globally if you configure EVPN in the default-switch instance, at the [edit multicast-snooping-options oism] hierarchy level.

• In the MAC-VRF instances if you configure EVPN in instances of type mac-vrf, at the [edit routing-instances instance-name multicast-snooping-options oism] .

We require that you configure install-star-g-routes with OISM on the QFX10000 line of switches, QFX5130-32CD switches, and QFX5700 switches when you configure those devices with AR in the AR replicator role.

In releases prior to Junos OS and Junos OS Evolved Release 23.4R1, you must also configure the install-star-g-routes option on the following devices when you configure them as OISM server leaf or border leaf devices:

• Switches in the QFX10000 line.

• PTX10001-36MR, PTX10004, PTX10008, and PTX10016 routers.

Starting in Junos OS and Junos OS Evolved Release 23.4R1, we no longer require you to set this option when you configure those devices as OISM server leaf or border leaf devices.

We don't recommend setting this option other than in the use cases mentioned above.

Consider this option only if you have very stringent latency requirements and can trade off higher scaling to achieve better network latency.

• Default Behavior without the install-star-g-routes Option
• Behavior with the install-star-g-routes Option

OISM and AR Scaling with Many VLANs

With OISM and IGMP snooping or MLD snooping enabled in an EVPN-VXLAN fabric, OISM server leaf and border leaf devices send EVPN Type 6 SMET routes into the EVPN core when their receivers join a multicast group.

When an OISM-enabled device receives Type 6 routes on the SBD, the device:

• Derives multicast states from the Type 6 routes as follows:

  • (*,G) states for IGMPv2 or MLDv1

  • (S,G) states for IGMPv3 or MLDv2

• Installs the derived states on the OISM SBD and revenue bridge domain VLANs in the MAC-VRF instance for all VLANs that are part of OISM-enabled L3 tenant VRF instances.

• Uses the derived multicast routes to optimize multicast forwarding by selectively sending the traffic for a group only to other EVPN devices that have receivers subscribed to that group.

On some devices that support OISM, you can also configure the assisted replication (AR) multicast optimization feature with OISM enabled. AR replicator devices use the Type 6 routes the same way OISM devices do.

QFX5130-32CD and QFX5700 switches can serve as OISM server leaf or border leaf devices. They can act as AR replicators only on devices that are not also OISM server leaf or border leaf devices. In that case, the device operates in the standalone AR replicator role.

The next sections describe configuration considerations on these devices when you configure them as OISM server leaf or border leaf devices, or as standalone AR replicators with OISM.

• ACX Series Routers, QFX5130-32CD Switches, and QFX5700 Switches as Server Leaf and Border Leaf Devices with OISM
• QFX5130-32CD and QFX5700 Switches as Standalone AR Replicators with OISM

PEG DF Election

By default, peer OISM PEG devices use PIM-based DF election—the devices discover their PIM neighbors on the OISM revenue VLANs and the SBD in each L3 VRF, and elect the DF from among those neighbors. Starting in Junos OS Release 23.4R1 and Junos OS Evolved Release 23.4R1, you can instead configure mod-based or preference-based PEG DF election using the peg-df-election statement at the [edit routing-instances name protocols evpn oism pim-evpn-gateway] hierarchy level, as follows:

When you configure PEG DF election, the device maintains an ordinal list of the PEG devices in the fabric that host each revenue VLAN (bridge domain) or the SBD. PEG devices use the EVPN multicast flags extended community in EVPN Type 3 IMET routes to advertise OISM, IGMP snooping, MLD snooping, and PEG device support. In addition, PEG devices include the DF election extended community to communicate the configured DF election method parameters (as defined in the RFC 8584 standard).

The peer PEG DF candidates are the devices that advertise IMET routes for a revenue VLAN or the SBD with the following EVPN multicast flags extended community values:

• For IPv4 multicast—igmp-snooping-enabled:oism:peg
• For IPv6 multicast—mld-snooping-enabled:oism:peg
  Note:

  The devices elect a DF for IPv4 multicast traffic and a DF for IPv6 multicast traffic separately based on those advertised multicast flags extended community values.

When the PEG devices use PEG DF election, they don't use the PIM protocol (they don't exchange PIM protocol packets) within the data center. As a result, we recommend you have external L3 redundancy on the PEG devices.

You can configure the PEG devices to use either of the following PEG DF election methods:

• Mod-based—The default method when you enable PEG DF election at the [edit routing-instances name protocols evpn oism pim-evpn-gateway peg-df-election] hierarchy level without including the mod option or the preference option. You can also explicitly configure the mod option to use this method.

  The algorithm to choose the device in the ordinal list that will be the DF is:

  (mapped VNI for the VLAN) mod (number of entries in the list)

  For example, if you have three peer PEG devices BL1, BL2, and BL3 configured with mod-based PEG DF election for VLAN 1, which is mapped to VNI 100:

  • The three devices each maintain an ordinal list of PEG DF candidates for VLAN 1 and VNI 100, such as:

    Table 7: Sample Mod-based PEG DF Election Candidates List

    Index	Device
    0	BL1
    1	BL2
    2	BL3

  • In this case, (mapped VNI for the VLAN) mod (number of entries in the list) is (100) mod (3) = 1, so the devices elect the device at index 1 as the PEG DF, which is BL2.

• Preference-based with a customized preference value—Configure the value preference-value option at the [edit routing-instances name protocols evpn oism pim-evpn-gateway peg-df-election preference] hierachy level. We recommend you set a unique preference value on each peer PEG device. With preference-based PEG DF election:

  • Each device communicates its preference value in the EVPN Type 3 IMET route advertisement.

  • The peer PEG devices elect the device with the highest preference value (by default) as the DF for the VLAN.

  • You can customize the preference-based method to elect the device with the lowest preference value instead of the highest value. To do this, set the use-least option at the [edit routing-instances name protocols evpn oism pim-evpn-gateway peg-df-election preference] hierarchy level.

With either PEG DF election method, you can also:

• Specify a time period to wait before the devices elect a DF. To do this, set the delay-time num option at the [edit routing-instances name protocols evpn oism pim-evpn-gateway peg-df-election] hierarchy level.

• Set the maximum number (0-255) of PEG DF election event entries the device maintains in a database for DF election history per VLAN (VNI). To do this, set the peg-df-election-history num option at the [edit <routing-instances name> protocols evpn] hierarchy level.

  The show evpn oism peg-df-status extensive command displays DF election history details.

All peer PEG devices must use the same DF election mechanism. As a result, if you enable PEG DF election, configure the same DF election method symmetrically on all peer PEG devices. If the configured PEG DF election methods don't match, the peer PEG devices all fall back to using the default mod-based PEG DF election method. If you enable PEG DF election on only some of the peer PEG devices, all of the devices fall back to using PIM-based DF election.