With the expansion of graph scale

single-machine algorithms can hardly adapt to the sub-graph queries in large-scale data sets. As existing distributed algorithms are based on simple traversal without index

the join process is prone to memory overflow in the distributed algorithms and load imbalance occurs when the query graph distribution is abnormal. Therefore

a binary index tree based on spectral coding named SCBT-index is proposed. Firstly

for vertex spectrum coding in the data graph

a binary index tree is constructed according to the coding information. Then

the query graph is decomposed using the minimum query plan. Finally

three pruning strategies are used in the join process: Structure matching based on topological structure

serialized join and the distributed join optimization. The experimental results show the comprehensive performance of SCBT-index under the graph set is better than that of the popular algorithms. In addition

the query time under the single graph is 1/2 to 1/4 of that of the existing algorithms.