In this paper, a new hash join algorithm-DHJ(the dynamic hash join) is proposed to resolve the problem of skewed data in the join operation in parallel database.The objective of the algorithm is to avoid the high cost of processing inherent in some early work. Additional buckets are used in the algorithm to balance output during the data partition. Then they are mapped to different processors before the completion of the join operation. The performance analysis of the algorithm is provided in this paper.