Abstract: The resource allocation issue in multi-services OFDMA systems is formulated by taking Quality of Service (QoS) requirement, data arrival process, system constraints into account. The relations among the traffic rate constraint, latency constraint and the queue length are analyzed. The original issue is decomposed into some independent sub-problems by dual decomposition scheme. The optimal resource blocks allocation scheme and optimal power allocation scheme can be obtained and a Dual Decomposition Based (DDB) optimization resource allocation algorithm is proposed. Simulation results show that total throughput maximum for non-real-time services is achieved by employing DDB algorithm with low service violation probability, nice fairness and slightly increased algorithm complexity.