An infrastructure model of communicating factories, warehouses, and sales outlets is considered. The facilities can be connected only rectangularly. Once the infrastructure network vertices are allocated, they should be optimally connected by fulfilling production and transportation services. This is solved as a mixed-integer linear programming optimization problem. The number of products which potentially can be manufactured by each of the factories is given as an integer. Each product is associated with a sales outlet via the demand, which is the quantity required to be sold through a definite period whose span can be left unspecified. Besides, manufacture capacity is given for each factory per product. Capacities of the warehouses are also known along with transportation costs, costs of manufacturing, and turnover rates. The goal is to minimize the total cost by ensuring that the demand is satisfied, while each sales outlet must be associated to only one warehouse. When the minimization problem does not have a solution, a practically applicable solution of an infeasible problem is obtained by successively increasing the manufacture capacity until a feasible solution is obtained, whereupon the amounts of products in the obtained solution are corrected so that the true manufacture capacity would not be exceeded. This method allows obtaining a solution of an infeasible problem at a rate of about 80% in the worst-case scenario of the infeasibility. Another suggestion addressing the problem poor scalability is to divide an initially large problem into a few smaller-sized subproblems by clustering facility locations, whereupon it is approximately solved by unifying solutions of the clusters.