With extensive applications of cloud computing, data capacity of data centers has grown rapidly. Furthermore, document information, which usually contains user's sensitive information, needs to be encrypted before being outsourced to data centers. Faced with such a large amount of ciphertext data, current techniques have low search efficiency in this scenario. Aiming at solving this problem, this paper proposes an efficient ciphertext search method based on similarity search tree (MRSE-SS) that can handle big data volume. The proposed approach clusters the documents based on the max distance between the cluster center and its members, constructs an n-dimensional hyper sphere by using the cluster center as the center of sphere and the max distance as radius, and then gradually clusters small clusters into large clusters. In the search phase of the ciphertext document collection constructed by this method, the ideal retrieval results can be obtained only by searching the query vector's neighboring clusters, thus improving the efficiency of ciphertext search. An experiment is conducted using the collection set built from the recent ten years' JC publications, containing about 2900 documents with nearly 4800 keywords. The results show that the presented approach can reach a linear computational complexity against exponential size of document collection. In addition, the retrieved documents have a better relationship with each other than by traditional methods.