Abstract: Nowadays it's desirable to generate traffic which can approximate realistic network traffic characteristics for performance evaluation of network equipments. To overcome the speed limit and accuracy insufficiency of existing traffic generation solutions, this paper proposes an efficient self-similar network traffic generator. The traffic generator can generate real-time self-similar time series according to the aggregation of multiple independent ON/OFF models. The main structure of the traffic generator is presented and implemented entirely using FPGA hardware resources. Statistical property analysis of the proposed traffic generator indicates that the generated traffic can achieve accurate self-similarity and better approximate the real network traffic. Moreover, the required date rates can be up to Gbps line rate and scale linearly with the increased port line rate.