A vSphere Distributed Switch is created at the data center level and spans across multiple participating hosts or clusters. Therefore, it can only be created by using a vCenter Server. The vCenter's data center inventory object is the boundary for dvSwitches.

A dvSwitch will have a single control plane and multiple data planes. The control plane is at the vCenter Server and is used to create and manage the dvSwitch. There will be a data plane created on each of the participating ESXi hosts. This means that all the packet switching will happen at the ESXi hosts. The use of a dvSwitch reduces the administrative complexity of configuring vSphere Standard Switches on individual ESXi hosts in a large environment.

Five different versions of vSphere Distributed Switches were available in this example: 5.0, 5.1.0, 5.5, 6.0, and 6.5. Choosing version 6.5 will make the vSphere Distributed Switch incompatible with older versions of the ESXi hosts (if they are managed using the same vCenter Server).

A dvUplink is another layer of abstraction added to reduce the administrative complexity. Every dvSwitch with dvUplink/s will have a dvUplinks port group. Every dvPortGroup created will increase the network count by one. Additional ports will always be consumed by the number of dvUplinks in the dvUplinks port group. This is true regardless of whether or not a dvUplink is backed by a vmnic from the participating ESXi servers.

dvSwitch configuration is saved in the vCenter Server database; however, a local host copy is maintained on every participating ESXi host at the /etc/vmware/dvsdata.db. The host copy is synced every 300 seconds. The dvsdata.db file being binary can only be viewed using the net-dvs CLI command. 

The dvsdata.db is only created after you connect an ESXi host to the dvSwitch. Running the net-dvs command without any ESXi hosts connected to the dvSwitch will yield no output.