Suppose there are $x$ servers, and $y$ users. The $y$ users are to be assigned to the $x$ servers similar to classic scheduling problems. The cost of using servers is given by $c(|x|)$ which is an increasing function of $x$, the number of servers. The more the number of servers used, the higher the cost function. There is another load function $L(x_i)$ which indicates the load, i.e. number of users assigned to each server $x$. The optimization objective is to Minimize the cost function of the servers, i.e. Minimize $c(|x|)$ such that the load at each server is within a pre-specified limit, i.e, $L(x_i) \leq k$.
If we look at the solution space of the problem, it comprises an exponential number of choices representing each user-server combination as per my understanding goes. What would be a reduction to prove the Hardness of this problem if my intuition about the solution space is indeed correct?