We propose to use a layered graph approach, which has been previously proposedfor unicasting, to have a more general, realistic and ï¬exible model of an all-optical mul-tiï¬ber network for multicasting. This new presentation enables us to state the problemof all-optical multicasting with sparse light splitting and wavelength conversion restric-tions so that it is formulated as an original Mixed Integer Linear Programming (MILP).The MILP formulation is solved by CPLEX which ï¬nds the optimal solution within agiven precision and it also gives a lower bound by relaxing the integrality constraints.However, it is possible to solve MILP problems to optimality only for small networksand number of sessions, since the problem is NP-hard. Therefore, we also propose threediï¬erent heuristics (LAMA, SLAM and C-FWA) for larger problems and dynamic mul-ticasting requests. Extensive computational experiments demonstrate that LAMA andSLAM perform close to the optimal and better than their competitor (M-ONLY) forall metrics. However, LAMA and SLAM work better than their alternatives, since wejointly optimize routing and ï¬ber-wavelength assignment phases compared to the othercandidates which attack to the problem by decomposing two phases. Experiments showthat important metrics are adversely aï¬ected by the separation of routing and ï¬ber-wavelength assignment. SLAM, which is the scalable version of LAMA, performs closeor better to LAMA. Finally, we also propose a new ï¬ber-wavelength assignment strat-egy (Ex-Fit in C-FWA) which uses wavelength and ï¬ber conversion resources moreeï¬ectively than the First Fit. |