Physical switches use MAC tables to perform frame forwarding. Physical switches learn and build the MAC table. The switch reads the Ethernet headers and learns the MAC address of the device and makes an entry in its MAC table.

Address Resolution Protocol (ARP) can discover a MAC address that corresponds to a known IP address, in order to permit the building of the layer 2 packets. From ESXi, you can see the ARP packets with this command:

[root@esxi2:~] tcpdump-uw arp
tcpdump-uw: verbose output suppressed, use -v or -vv for full protocol decode
listening on vmk0, link-type EN10MB (Ethernet), capture size 262144 bytes
15:28:23.878389 ARP, Request who-has 10.10.70.110 tell nsxmanager.lab.local, length 46
15:28:23.879198 ARP, Request who-has esxi2.lab.local tell labserver.lab.local, length 46
15:28:23.879316 ARP, Reply esxi2.lab.local is-at 44:a8:42:22:a3:f2 (oui Unknown), length 28

On ESXi, there is a local ARP table for all the VMkernel interfaces, and you can print it with the command esxcli network ip neighbor list.