With the continuous expansion of network scale, evaluating the performance of IP network can help guide network design and provide support for improving the operation of the network. Aimed at the shortcomings of the fussy method and projection pursuit method, a new grey fussy evaluation method is put forward. Five key indexes to influencing the network performance are selected firstly, and the correlation degree of each evaluation index is obtained by using grey relational analysis, which is used as index weight set through normalizing. And the fussy relationship matrix of index set to the comments set is constructed after the fussy calculation is performed, finally the performance evaluation and the ranking of the seven networks are completed. Compared with grey correlation, TOPSIS and projection pursuit method, the result shows that the inner relationship of index is reflected in the results of evaluation, the algorithm is more objective in determining the weight and the specific value of network performance is given. The algorithm is simple, and provides a new idea for the network performance evaluation.