I would consider placing an Agent Handler (AH) and Remote Repository in that location. Set the machines there to communicate only with the AH. Then you would only have the communication from the AH back to the ePO server. The machines could also be set to use the local Remote Repository thereby reducing the hit to your bandwidth further.
Thx for the suggestion. Thats how my setup is.
Adding to my question.
I have enabled RSD task of deploying the agent once the rogue is found. So, ePO server pushes the agent to the rogue from ePO Server itself, not from the remote agent handler/DR.
My question is whether there is any other alternative for this as this consumes bandwidth.
In the server task that you created to push the agent on Rogue machines you can specify to select Agent handler Under 'push agent using=>selected agent handler'.