Abstract: Graph data models are widely used in various area to present, store and process the data with complicated relationships. Considering the deficiency of existing music data models and query languages, this paper firstly presents a graph-based music data model named Gra-MM to model music data with complicated relationships. The definitions of model’s logical data structure and algebraic operations are given. Then based on Gra-MM, we present a music data query language called Gra-MQL, with the BNF syntax of the language. Gra-MQL can handle the complicated relationships among music data well, and also has the ability to query music by meta data as well as music content data. Gra-MQL meets a variety of user requirements and overcomes the shortcomings of traditional graph query languages which do not have adequate expressive power when dealing with data that have complicated relationships and don’t have the ability to query music based on music content data. Finally, a brief introduction on the prototype of music database management system is given in this paper. Three queries experiments are carried out on this prototype and the performances on different datasets are given, and these experiments show that the model and query language are both feasible in practice and theory.