7 Case Study and Conclusion Bws(fwsg,fwsg) Bws 8w 2 W and 8s 2 S Solution via Simulated Annealing 6 Utilization Approach

In this section we shall present the results obtained by applying the two formulations proposed in this paper to a case study. With respect to the service characterization, we considered a network ooering only three diierent service types which we found reasonable to characterize a multiservice network while reducing the amount of data to be presented here (Table 1). Service Holding Time Connections s (1/s) (N max s) 1 100.0 3.0 2 20.0 10.0 3 5.0 50.0 On table 2 we present, for the call blocking minimization approach, the available capacity in each ber trunk, the amount of ber capacity allocated to second order arcs and the remaining capacity left idle on the ber trunks. Note that with this approach there is a tendency to allocate all the available bandwidth as predicted earlier. On table 3 we present the corresponding results produced by the utilization minimization formulation. As it is implied by its name, with this formulation only the strictly necessary capacity is allocated in the ber trunks as can be seen on the table. One can note that now we have plenty of idle capacity on the ber trunks. From a practical implementation point of view where the dynamic reconnguration must be processed in real time, the utilization minimization approach is much more eecient. For the case study presented, the call blocking minimization approach took some 7 min to reach its results while the utilization approach took only 17 s. In both cases, the algorithms were excecuted on a SPARC Server 630MP. Hence, from an eeciency point of view the utilization minimization approach is more likely to be chosen for a real implementation. Table 3: Fiber trunk utilization produced by the formulation in section 6. \An eecient near-optimal algorithm for the joint traac and trunk routing problem in self-planning networks", in Proc. X ap2Pw ps = ws 8w 2 W and 8s 2 S (6) M X u=1 C au :p u C (7) C 0 (8) where, C = (C1; C2; ; CM) T is the rst order arc capacities vector, C = (C a 1 , C a 2 , , C a M) T is the second order arc capacities vector and = P w2W w. The minimization in (4) takes into account the call blocking probability imposed to every service s 2 S traac ooered to every node pair w 2 W. The constraint (5) …