Abstract: The contradiction between the processing of mass network traffic data and the computing bottleneck of a single node leads to low efficiency of data classification. To address this challenge, we propose an ontology based parallel network traffic classification method by integrating the advantage of ontology and MapReduce in dealing with the description and processing of mass heterogeneous data. Our approach makes use of MapReduce, a framework of parallel computing. Firstly, it uses the ontology to describe and manage network traffic data, and constructs the layered and parallel network traffic ontology. Then it builds the classification model by employing the decision tree algorithm, by which the inference rule set is generated. Network traffic classification based on traffic statistical features is completed by utilizing parallel knowledge reasoning. Implementation results show that data classification efficiency of the proposed approach in group environment is higher than in stand-alone scenario. The speedup ratio increases linearly when increasing the quantity of compute nodes. In addition, the new method is able to improve the classification efficiency of large-scale network traffic significantly.