Head end mode is flood learn concept.Source Hypervisor does the flooding of the tunneled packets which ultimately result in more bandwidth usage and some additional overhead on ESXI host.
Two Tier Mode - Instead of sending X(based on the host count) amount of tunneled copied, it will send few copies ( Based on the TEP subnet group-1 per group) which will result in less bandwidth utilization and less overhead on esxi host so this will be a better option irrespective of L2/L3 fabric .
Thanks for the reply, the answer helped me.
If i'm designing an infrastructure when can I use the two tier or Head End. Since the two tier can do the same job as the head end.
Are there specific use cases where it is preferable to use the Head End rather than two tier?
Head end is default. If your TEPs are all on the same L2, for example, it makes no difference which one you choose. Only if your TEPs span across L3 would you maybe gain some performance improvement with two-tier.
I don't find any real advantage in head-end replication . In virtualization ,we all prefer load sharing and balancing for Compute/Storage/Network . So Two-Tier is certainly preferred when TEP's are in different subnet. In a single Subnet TEP , irrespective of the option what is selected, there is no difference - It will be a just a cherry picking check button
Thanks for your help. I thought two-tiers was the default