Intermittent snapshot method for data synchronization to cloud storage

We developed a configuration to place a cash storage called Cloud on-Ramp (CoR) with small capacity between client terminals at remote offices and a cloud storage connected by a wide area network. A CoR is placed in each remote office, and applications access them via a local area network, guaranteeing access performance. A CoR synchronizes the data with the cloud storage by copying updated data from the client terminal at regular intervals. It is necessary that a window to carry out the bulk-copy function be shortened and repeated many times to keep the latest data in the cloud storage. A method of maintaining consistency of data at a CoR is necessary, even if the data are updated in the copy by the client terminal. Therefore, we propose the "intermittent snapshot method", with which a snapshot is taken during bulk-copy execution and released as soon as the bulk copy is over. We evaluated the proposed method from the perspective of an implementation design. We formalized the method by using a stochastic Petri-net model and considered the proper size of the bulk-copy window that optimizes both the synchronization delay and application-access performance through simulation.