Use case of BUM Traffic?

[MrVmware9423]

Dear Team,

Why do we need BUM traffic if we are already keeping TEP, MAC, and ARP tables both locally and centrally?

.Please throw some light on it.

Thank you in Advance.

[Sreec]

Well, the answer is in the question :). BUM (Broadcast, unknown unicast, and multicast) must be flooded to all workloads in the same segment, which can be on the same host or a different host. NSX replication modes handle this with head end or two-tier replication modes. In a nutshell, not in all cases, ARP will be resolved locally or centrally

[MrVmware9423]

That's what, I would like to know in what all cases BUM traffic will come into the picture? as we are already maintaining TEP,MAC & ARP table....

[Sreec]

Just giving one example, ARP tables will not be populated upfront. Whenever an ARP request goes through, the initial packet will be flooded, however, tables will get populated eventually and it will reduce further flooding. Layer 2 broadcast, unknown unicast, and multicast traffic will be flooded. This is not specific to NSX, this is common in any network however different vendors have their own replication modes. In NSX we offer two modes 

1. Hierarchical two-tier (sometimes called MTEP)
2. Head (sometimes called source)