In this paper, we concentrated on resource allocation for heterogeneous traffic in OFDM wireless networks. We considered three types of traffic, i.e. hard QoS, soft QoS and best effort (BE), the characteristics of which can be described by three different utility functions. We proposed a utility-based cross-layer resource scheduling model, which performs combinational-optimization adaptive resources allocation. At first, the scheduling model is formulated as a non-linear integer programming problem with the objective of maximization total utility, constrained to the CCI restriction and the quality-of-service requirement of QoS traffic. And then, it's transformed into a continue relaxing convex programming problem and solved. Based on the optimal relaxing solution, a low complexity dynamic subcarrier allocation algorithm, namely MHSU algorithm, is provided. Simulation results demonstrate that our scheduling algorithm guarantees a good performance of supporting heterogeneous services, and the system utility value obtained by MHSU algorithm is very close to the optimal value.