Since you seem to refer to a network inside a virtual host:
The physical topology is strictly point-to-point (nowadays), but the topoplogy on the data link layer is point-to-multipoint: each node can send a frame addressed to any other node out of (one of) its interface(s), and the network takes care of that frame reaching its destination.
Usually, a switch forwards the frame to the correct destination indicated by its destination MAC address. At the same time, the switch learns by the frame's source address on which port the source host is located and updates its internal MAC table.
You can also simply connect multiple switches to each other, forming a larger network. The only difference is that a switch now might learn multiple MAC addresses on each port.
A virtual host connected to a physical switch acts just like a switch itself: it makes the physical switch learn each MAC address from any of its virtual machines and the physical switch forwards frames in the correct way.
The host's NIC responds to many MAC addresses and the host's virtual switch then forwards them logically to their respective vNIC connected to one of its VMs.
A VM in turn can be an end node - consuming the frame - or a switch/bridge itself, forwarding the frame out of another one of its vNICs, thus bridging different port groups.
A router (on the network layer) is also an end node from the perspective of the data link layer - it consumes the frame. However, it does not consume the contained IP packet but de-encapsulates it and then forwards it, according to its routing table, out of one of its interfaces (re-encapsulated in a new frame).
(I have simplified somewhat - the data link layer doesn't necessarily use MAC frames and the network layer not necessarily IP packets.)
