Distributed switches are required.

Configure the MTU to be at least 1600 at ESXi and on the physical switches. This avoids IP fragmentation.

Map the guest MTU size to accommodate the VXLAN header insertion at the ESXi level.

Use explicit failover or “route based on IP hash” as the load balancing policy.

Multicast - If VXLAN transport is traversing routers, multicast routing must be enabled (PIM – BIDIR or SM).

More multicast groups are better.

Multiple segments can be mapped to a single multicast group.

If VXLAN transport is contained to a single subnet, IGMP Querier must be enabled on the physical switches.

Use PIM-Bidir (so that any sender can be a receiver as well) where available. Otherwise use PIM-SM.

If VXLAN traffic is traversing a router, proxy ARP must be enabled on first hop router.

Leverage 5-tuple hash distribution for uplink and interswitch LACP.