From what you describe here, I would think a distributed repository (of any kind - not just a superagent) is not required. The time you would start to need a distributed repo would be if the number of client machines updating from the master repo was adversely affecting the ePO server itself.
Your reply helps me out although it did not answer my question regarding the impact of deploying SADR in case it is mandatory to have SADR.
For example, one of my subnest has only 2 clients. What would happen if I don't have SADR on that subnet while having SADR on the rest?
Below is the quote from McAfee KB.
To deploy sufficient numbers of SuperAgents to the appropriate locations, first determine the
broadcast segments in your environment and select a system (preferably a server) in each to
host a SuperAgent. Be aware that agents in broadcast segments without SuperAgents do not
receive the broadcast wake-up call, and therefore, do not call in to the server.
Super Agents basically enable two functions:
- Distributed repository (glorified file share)
- Broadcast wake-up call (minor traffic-saving function)
The broadcast wake-up call only saves on out-bound traffic from the ePO server--NOT the return traffic, which is the substantially larger traffic anyway.
In your case, the ePO server would direct your single Super Agent to broadcast to it's segment, all the agents in that segment would then Collect&Send Props back to ePO, and any uncovered segments would blissfully continue (not communicating) until their regular communication interval rolls around.
Most people just send a normal, direct wake-up call to the system they care about. 99% of the people never even know about the broadcast wake-up, and that's the part where the segments matter.