As a data-intensive computing environment

data grids require not only high performance computing resources

but also timely high speed access to the data resources

which are needed at the computational resources.The performance of the storage accesses is key to the performance of data grids platforms.The paper presents the idea of dynamic storage hierarchy appropriate for the data grid environment

and puts forward an approach of its construction by grid locality aggregation.The paper also explores the dynamic K-clustering feature of the approach

tests and analyzes the performance of the dynamic storage hierarchy.Experiment results show the explicit improvement of the overall grid tasks handle based on the dynamic storage hierarchy.