Data retrieval from multiple heterogeneous cloud databases is a complex engineering task. Ensuring scalability, flexibility and data availability while integrating data from the multiple heterogeneous cloud databases are the major research issues in such task. A proper integration semantics for cloud databases and data retrieval through Database as a Service (DBaaS) can be helpful to solve those issues. In this paper, a graph semantic based model for SaaS level database integration (SLDI) has been proposed, which is useful towards realization of various cloud database integration approaches through services. Several important quality metrics like integration complexity, degree of consistency, degree of availability and degree of partitioning tolerance also have been defined over the proposed model. The proposed SLDI model is compatible with CAP (Consistence-Availability-Partitioning) theorem. Besides, the proposed approach is capable of handling data heterogeneity through service while ensuring scalability and flexibility in data retrieval from multiple cloud databases.