Intermittent snapshot method for remote data synchronization from cloud on-ramp to cloud storage

To avoid low data-access performance through wide area network (WAN) from clients at remote offices to a cloud storage at a data center, we proposed a cloud-storage cache, called "cloud on-ramp" (CoR). Because the CoR is located at the remote office and connected to the clients through local area network (LAN), it can improve access performance from applications. However, a bulk copy that synchronizes data from the CoR to the cloud storage causes performance overhead. Especially, the more periodical bulk copies occur, due to ensure that distributed users can share more fresh data through the cloud storage, the more snapshots need to be taken, which keep data consistency but cause performance degradation for the applications. To ensure that the applications have reliable access to data even in the case of the periodical bulk copies, we propose a snapshot method called "intermittent snapshot" and a remote data synchronization method based on it. Execution timing of both the bulk copy and the snapshot completely synchronize, and the snapshot terminates after the bulk copy completion during the remaining period of a periodical bulk copy window, in which the application access performance ensures. In addition, we evaluate the remote data synchronization based on the intermittent snapshot from the perspective of an implementation design. We formalize the method by utilizing a stochastic Petri-net model, and determine a proper size of bulk copy window that optimize both synchronization delay and application access performance through Petri-net simulation. We also evaluate its data access performance improvement in comparison with a conventional method and a limit performance of the method.