In a large distributed system, it is important that computers share the loads submitted to the system in order to harness all its computing power. This paper proposes and analyzes a new algorithm for sharing system loads based on a coordinated dynamic semi-centralized scheme. In this scheme, load state information at each node is centrally collected by a coordinator to optimize the transfer policy. This information is periodically broadcasted to nodes for decentralizing the decision, making for best load allocation policy. The paper analyzes the algorithm for small systems as well as for large distributed systems. Simulations results have shown that this semi-centralized scheme outperforms probing-based schemes in most system loads.