VXLAN Layer 2 Packet Walk (BUM Traffic)
VXLAN EVPN BUM Traffic Packet Walk

VXLAN Layer 2 Packet Walk (BUM Traffic)

Last topic we discussed the bridging traffic for known host and explored how traffic is handled and what kind of route type is sent between leafs and each other.

But what if the traffic is not known unicast, and it was broadcast, multicast or unknown unicast, how is traffic is handled in this case? This is what we will cover in today's topic.

Will start again with our dataplane mindmap:

Article content
VXLAN EVPN Data Plane

As clear above we have two types of forwarding BUM traffic:

  • Ingress replication (Unicast replication)
  • Multicast

Let's Start with Ingress Replication (IR):

In ingress replication send unicast packet to all participants in the VNI.

which means that when Leaf receive BUM traffic it will send unicast packets to each leaf in the network who is participating in that VNI.

The question here is:

How the leaf to who to send the BUM traffic?

The answer is, because of EVPN, one each Leaf device is configred with the corresponding VNI, so it will send to all discovered peers that i'm configured with that specific VNI and here it is my Next-Hop its VTEP IP.

EVPN is responsible for transporting this data between leafs using RouteType-3 as per below:

Article content
EVPN NLRI Route Type-3

After each leaf share send its EVPN NLRI route type-3, So each leaf now has its own flooding list table:

Article content
Ingress Replication Flood List Tables (Forwarding Table)

So, now we know how each Leaf will send its traffic and who it build its forwarding table.

Now Leaf-1 (L1) as an example will start sending unicast packets to all leafs in its flood list table as shown below.

Article content
Host Send BUM traffic

And after each received the VXLAN packet it will de-encapsulate and forward it according to its local table (MAC and ARP) table.


BUM Traffic (Multicast):

In IR (ingress replication) we knows that the source leaf will send unicast packets to all leafs in its flood list table.

Also we know how it build its flood list table

****In send BUM traffic using Multicast option the control plane is different than IR in building its fl

Let's assume the below topology and Host A need to send broadcast traffic, So it send its packet to L1 and L1 here need to send it as well to all Leafs joined in same VNI.

So, again the question here is how Leaf-1 knows to which leaf i should send the BUM traffic.

Now in Multicast option, the building mechanism of flood list table is different than IR.

As in this scenarios we are configuring the mapping of VNI to which multicast group.

Which means the flood list table is different than in IR as shown below:

Article content
Flood List table in Multicast Scenario

So Leaf-1 will encapsulate the received traffic from Host-A and destined it to multicast group as shown above.

In this scenario the spine is the RP for the multicast tree, So Spine will replicate the packet to all leaf registered in the multicast group.

Article content
Host-A sends BUM Traffic using Multicast scenario

Conclusion

  • We discussed that there are two options to send BUM traffic (IR or Multicast)
  • We discussed how BUM traffic is sent using IR option and how each leaf build its flood list table
  • We discussed how BUM traffic is sent in Multicast option and how flood list table is built based on multicast group configred per VNI.

To view or add a comment, sign in

Others also viewed

Explore topics