Most of the existing studies on traffic grooming focus on minimizing required network link capacity and providing a serving relationship between client services and link capacity. Subsequent to this step, it is important to plan for actual client service add/drop over client service ports and end-to-end lightpath establishment over network ports, which is, however, not well investigated. We call such an effort node hardware module planning. This is an industrially practical problem aiming to minimize the node hardware cost since hardware modules are usually the most expensive in a network. Based on a link-based traffic grooming result, we develop a mixed integer linear programming (MILP) model to optimally plan hardware modules. To overcome the computational difficulty of the MILP model under large-size planning scenarios, we also develop a fast suboptimal heuristic for hardware module planning. Simulation studies indicate that the heuristic is efficient to realize a design close to an optimal solution obtained by the MILP model for both of the single-hop and multi-hop grooming modes. Also, the multi-hop grooming mode requires not only fewer link capacity units than the single-hop mode as found in most of the existing studies, but also lower node hardware costs. Finally, the evaluation of the impact of the switch backplane size shows that given a certain set of hardware modules, a saturated switch backplane size exists after which a further increase of the backplane size will not bring further reduction of the network hardware cost.
© 2011 OSAFull Article | PDF Article