The load balancer will use an algorithm to split the users connections to the front end servers. You don't assign users to a specific front end server. The users could potentially be given the url for a specific machine name if they were
using the web client to connect but you would likely see issues. Especially with custom code. And it also defeats the purpose of load balancers.
The same with the back end server. The load balancer will assign connections for an async service to connect and grab jobs from whatever backend server has the least workload based on the built in logic of the load balancer. The async service grabs
a specific number of jobs to process each time. So if the first server grabs the first group, and there are not that many left, theoretically the second back end server could have a lighter workload.
So you need to monitor the workload to determine how many of each front end and back end need to be added. Each environment is unique and defendant on hardware.
Jason Peterson