There are two models in which you can design VLANs on a network:
End-to-End VLANs are dispersed throughout the network on multiple switches. No matter where the user plugs in, they will always be assigned the same VLAN membership. Resources for this user will be on the same subnet. Management of end-to-end VLANs typically use VTP through trunked switches. This allows for easier management.
On the other hand, VTP can become a nightmare. Additionally, because the VLAN is spanned across your network you allow broadcasts to traverse multiple switches. This model doesn’t scale very well for large environments but is okay for small implementations.
VLANs are localized to a single switch. Usually based on geographic location. If a user moves to a different location, they will get associated with a different VLAN.
In the Local VLAN model, the local switch is connected to a distribution or core switch. This is where the routing is performed. Local VLANs will only exist between the local (access) switch and the distribution or core switch.
To reach other networks, traffic is sent from the local VLAN to the distribution/core switch and then routed to its destination. In this model, VTP is not configured and switches operate in VTP Transparent mode. VTP is no longer needed because the VLANs are not being advertised to other switches on the network.
Benefit of this model is you eliminate extended a layer 2 broadcast domain across your network. Troubleshooting is simplified, in comparison to end-to-end VLANs, because traffic will flow a certain path. End-to-end VLAN traffic flows through multiple switches.