Flexible replication for personal clouds

People own an increasing number of personal devices ranging from mobile phones and laptops to tablet and desktop computers. In addition, it is more and more common to rent cloud storage resources from utility computing providers. We call this new computing environment a user’s Personal Cloud. Managing data in such a heterogeneous environment requires a large effort on the user side. Given the storage limitations of some of these devices and the increasing amount of digital information stored on them, users demand a way to partially replicate their data across the devices in their Personal Cloud. Such a replication system must be flexible enough to fulfill the requirements of various potential applications. The heterogeneous devices may have different communication capabilities which disallow direct communication to every other device in the Personal Cloud. Furthermore, disconnections and failures of the devices must also be taken into account. In this thesis, we present a flexible replication infrastructure that is inspired by the PRACTI system and offers partial replication, arbitrary consistency, and topology independence. We have integrated our system in the Rhizoma runtime to benefit from its self-monitoring functionality and the ability to dynamically redeploy in the execution environment. We have extended PRACTI with metadata-based grouping of content and support for content-based filters. We have assessed the performance of our replication system on a realistic testbed consisting of mobile devices, desktop PCs and cloud virtual machines. Results show the feasibility of our approach and the incurred overhead.