An algorithm based on double counter bloom filter for long flows identification (CCBF) is proposed in this paper. Double counter bloom filter structure is used to distinguish the process of the long flow filtration from the long flow existence. The false positive rate of the algorithm is analyzed. The relationship of the memory requirement and the error rate is analyzed through simulation. It is shown that with the same restriction of the memory resource, the average error of this algorithm is less than the existing similar algorithms. The analysis of the time performance shows this algorithm is capable of dealing with traffic up to 1 500kpps.The results reflect that this algorithm can be used to monitor the long flows on backbone network.