As known, it is difficult for the multiple constraints QoS routing to get a polynomial-time solution. Currently, many heuristic algorithms based on the path distance have been proposed to solve this problem. However, most of them are based on linear path distance, which has the greatest shortcoming, i.e., the search of feasible solutions is inefficient. Besides, they don?t also consider the link sharing interference and load balancing. To improve the performance, this paper proposes a novel QoS routing based on the nonlinear path distance, which is called MIS-LB. At first, the additive and multiplicative QoS metrics are represented by a uniform metric, i.e. the nonlinear path distance. Then, the link sharing and load balancing are integrated into the uniform metric. Based on the shortest-first criterion, MIS-LB can not only find feasible paths as the normal routing algorithm (e.g., Dijkstra algorithm), but also adjust the link sharing, and balance the network loads between multiple feasible paths. So compared with TAMCRA and H_MCOP, the simulation results show that MIS-LB can further improve network performance.