A new assignment algorithm for star network topology design

To solve the star topology communication network design problem, the following optimization tasks must be solved: the concentrator quantity problem, the concentrator location problem and the assignment problem. The two last mentioned problems are NP-hard and the optimal solution of these problems has not been found for large networks. Algorithms looking for the optimal solution of the assignment problem (with minimal total link cost) can be used only for small networks with several hundreds of terminals. The assignment algorithm presented in this paper finds a superior solution in a reasonable time also for large networks with a few thousands of terminals. The computational results show that the algorithm is superior to "simulated annealing" and "tabu search" not only in CPU time comparison, but also is characterized by its lower gap to the optimal solution.