A Bi-objective integer linear programming model for the routing and network coding assignment problem in WDM optical networks with dedicated protection

The explosive traffic growths are pushing the transport network close to its capacity limitation, raising critical issues about fiber capacity crunch. In this context, network coding has been emerging as the promising technique to improve the network capacity efficiency thanks to the capability of better resources utilization. The application of network coding to the realms of failure recovery in optical networks has paved the new way for more efficient protection schemes and indeed, XOR network coding combined with dedicated protection has been proposed, investigated and developed to challenge the well-established understanding of trading capacity efficiency for recovery speed and vice versa. In order to maximize the benefits empowered by network coding in this case, the problem of 1 + 1 routing and network coding assignment (1 + 1 RNCA) has to be optimally solved. Apart from traditional 1 + 1 routing, the decision of network coding information has also to be taken into account including the selection of pair of demands for encoding and the respective coding node and coding links. In this paper, we propose a bi-objective integer linear programming model of the 1 + 1 RNCA problem aiming at minimizing the conventional routing cost as the primary objective and furthermore minimizing the number of nodes with coding capabilities as the secondary objective. Our formulation uses a weighting method to combine two objectives into an integrated one and we provide a rigorous analysis on configuring the weight coefficients to capture the desired priority of individual objectives. The efficiency of our integrated objective model in comparison with reference designs based on the single-objective model, 1 + 1 routing and 1 + 1 RNCA, is numerically evaluated on different realistic topologies and traffic sets. Extensive simulation demonstrates that our proposal outperforms traditional approaches when it could achieve the lowest routing cost while simultaneously employing minimal number of coding nodes, albeit with a slightly higher computation time. We furthermore provide a parametric study on the impact of number of coding nodes to the achievable routing cost improvement and it turns out that having two coding nodes is already sufficient to realize most of the benefits enabled by network coding.