i usually segregate the servers, services and devices. (mind you we have thousands of devices registered to multiple clusters)
for example, i have 4 different VLANs for the ccm servers in multiple clusters.
pub has it's own vlan, tftp has it's own vlan and subs are divided among two other vlans for network redundancy reasons. (if one vlan dies the phones/devices still have access to a sub in another vlan)(i also have two of the subs as redundant tftp incase the main fails)(i have redundant data centers at the core & dist. with redundant/meshed uplinks to the accesss layer so i'm up even if one of my datacenters loses power; must be nice to have $$$$ to spend!!)
the publisher does nothing but CDR, administration, authentication and web serving, ie: directories, services, etc.
the use of multiple VLANs in this scenario allows me to make some specific redundancy as well as segregate traffic easier via VACLs. (vlan acl's)
we are extremely buttoned up.
no access to any services/devices unless expressly permitted, no matter where on the network, egress & ingress. (i don't even let a device talk to other devices on its own VLAN unless permitted; using the most acl's i've ever used) you may not need so much security in place, hence you may not need some of the benefits of diseminating the servers/services across the network as such.
if you ask a well trained/experienced network designer, i'd think you would get an answer of segregation for security purposes, and consolidation for ease of administration and troubleshooting purposes.
