Abstract: The sequence suffix is partitioned into several parts by using appropriate policy, each partition is processed independently on the cluster systems to produce a temporary result, the final maximal tandem repeats set is achieved from the union of all temporary results, and a parallel algorithm for long sequences maximal tandem repeats recognition is presented based on partitioning suffix array. The parallel algorithm can decrease processing time and space consumption. The experimental results also show that the algorithm is flexible and scalable.