This paper considers the optimization problem of traffic allocation in a connection-less autonomous system.The objective is to enable each node in the network to independently distribute traffic among any given set of next hops in an optimal way.We model the network utility function with non-concave function because of users’ inelastic requirements on data rates in some practical applications,such as video and audio streaming.Therefore,we tackle the problem of optimizing a generalized class of non-concave utility functions.To solve this problem,we first design a sequence of convex relaxations whose solution converges to that of the original problem.Then,we propose a distributed sending rate control algorithm to solve the convex relaxation problems.The approach taken relies on results from real-analysis theory and optimization theory.All computations required by the algorithm are performed independently and locally at each node using local information and minimum information exchange between neighboring nodes.Numerical simulations are performed to evaluate our results.