This paper considers a general design problem for a parallel processors system which consists of a set of non-homogeneous potential processors and a set of traffic distributing points. For this problem, two levels of technology choice are made: 1) which candidate processors to establish with capacity option, 2) how to allocate the incoming traffic among established processors generated at each distributing point. The associated model is formulated as a linear fractional integer program whose objective is to minimize the maximum mean queue length on any established processor, while satisfying a side constraint such as budget constraint. Within the framework of branch and bound procedure, fractional subproblems are optimized by solving a sequence of linear programming problems. Computational experience with randomly generated 25 test problems is presented.