greetings...
the basics + rules of thumb...
1 data collector per 200 targets (full assessment)
1 data collector per 500 targets (basic assessment)
collectors can run across WANs - however, one collector per location is recommended
Data Manager has to be installed as local admin
then... for the discovery...
You can decide whether to perform a discovery OR import a csv file
If the client has an accurate listing of their servers, i like the import method.
If not, do a limited discovery - as quite often, customers find out about servers they never knew they had 
ie. server/desktops sitting under peoples desks or hidden away in back rooms.
here are some pointers...
On Windows systems - ports 135, 137 through 139,and 445 need to be open
On Linux and UNIX systems, port 22 is required for Secure Shell (SSH)
Trusted Domains - The collector host does not have to be in a trusted domain. However, the fully qualified account the collector uses for connection to the target systems must be one of the following:
Account of the target system
Account of the target system domain
Account of a trusted domain of the target system
Account Privileges
Windows: Inventory and performance data collection through WMI, Remote Registry, and Perfmon. An account with local administrative rights to the target system required.
Linux and UNIX: Inventory and performance collection through SSH scripts. A root equivalent administrator account is required.
hope this helps