Fast auto recovery of data, efficient copy management and a balance dependability of replicas and performance of cluster server can be achieved by this system scheme, which is proven by performance test of the disaster recover system.