Swiftmend: Data Synchronization in Open mHealth Applications with Restricted Connectivity

Abstract

Open mHealth applications often include mobile devices and cloud services with replicated data between components. These replicas need periodical synchronization to remain consistent. However, there are no guarantee of connectivity to networks which do not bill users on the quantity of data usage. This thesis propose Swiftmend, a system with synchronization that minimize the quantity of I/O used on the network.

Swiftmend includes two reconciliation algorithms; Rejuvenation and Regrowth. The latter utilizes the efficiency of the Merkle tree data structure to reduce the I/O. Merkle trees can sum up the consistency of replicas into compact fingerprints. While the first reconciliation algorithm, Rejuvenation simply inspects the entire replica to identify consistency. Regrowth is shown to produce less quantity of I/O than Rejuvenation when synchronizing replicas. This is due to the compact fingerprints.