Data Grid is one of the important research subjects in the field of Grid Computing. To get a high efficiency of the system, this paper proposes a set of novel multi-cast algorithms for Data grid, which can achieve shorter multicast delay and less resource consump-tions. In contrast to previous approaches, the algo-rithms in this paper partition the group members into clusters in the lower layer, seeking the most plenty of date amount node as an optimal core (root) and build-ing a shared tree within each cluster according to the algebra addition of the date amount of node and expo-nential to minimize the number of links used. In the upper layer, a shared tree is then constructed using our algorithms to implement the inter-cluster routing. The extended simulation results indicate that our algo-rithms are more suitable for Date Grid as compared with other well-known existing multicast solutions.