Optimization of voice and data routing in a burst-switched network was handled by means of linear programming multicommodity flow models. Routing was assumed to be source independent, with random bifurcation, and allowed nonuniform freezing of voice requirements. It implicitly assumes global knowledge of network steady-state statistics, and does not account for adaptive rules. The authors present numerical solutions of the resulting linear programs in small sample networks and discuss parameter tuning that produces reasonable compromises among the conflicting performance objectives. They include all conflicting multiple objectives and constraints in a linear programming formulation and show how parameters can be tuned to produce desirable voice and data paths.