【摘要】将来网络服务质量需要考虑网络业务流自有的自相似性对网络性能的影响，而传统的网络模型未考虑此因素。该文利用分数布郎运动模型，建立了一个网络模型，此模型考虑了业务流的自相似，以及聚类的特性，通过控制有效带宽，提高统计复用效率，并通过审核控制机制，提供网络服务。实验证明分数布郎运动模型能如实描述业务流特性，准确预测业务流的性能，能够最大限度地利用网络资源，提高服务质量，弥补原来网络排队模型的缺陷。关 键 词 自相似业务流; 服务质量; 准许控制; 分数布朗运动Network quality of service in future needs to consider the impact of inherent self-similar. Traditional networks ignore this. The paper utilizes Fractional Brownian Motion model to build a network model, which improves statistical multiplexing efficiency by controlling effective bandwidth and provides services by admission control. Experiment verifies that FBM model can realistically describe packet traffic flows，predict their performance, make full use of network recourses, and make up the deficiency of conventional queue models.