This paper studies a version of the fixed-charge multicommodity network design problem
where in addition to the traditional costs of flow and design, congestion at nodes is explicitly
considered. The problem is initially modeled as a nonlinear integer programming formulation
and two solution approaches are proposed: (i) a reformulation of the problem as
a mixed integer second order cone program to optimally solve the problem for small to
medium scale problem instances, and (ii) an evolutionary algorithm using elements of iterated
local search and scatter search to provide upper bounds. Extensive computational
results on new benchmark problem instances and on real case data are presented