Restorable logical topology in the face of no or partial traffic demand knowledge

The construction of a logical network on top of a physical (optical) infrastructure involves two intertwined tasks: logical link selection - deciding which pairs of routers will be connected by logical links (lightpaths), and logical link routing - deciding how to route each logical link across the optical network. The operator of such networks is often required to maximize the available throughput while guaranteeing its restorability. This paper is the first to combine these seemingly conflicting goals into one optimization criterion: maximizing the restorable throughput of the end-to-end paths. We address this problem in three cases: when the operator has no knowledge of the future bandwidth demands, when it has partial knowledge, and when it has full knowledge. We present efficient algorithms for each of these cases and use extensive simulations to compare their performance.