As computer users, we create increasing amounts of data, such as digital documents, pictures, and videos. Because these data have high value in our daily life the need for back-ups arises. The creation of local back-ups, e.g. on external hard drives, is a common approach, but is insufficient in the event of natural disasters or theft.
In this thesis, we provide a proof of concept for a distributed back-up system that induces only low overhead, and respects user needs to easily recover a state of a file system in a snapshot-based manner. We store distributed back-ups on residential gateways and use a central tracker as coordinator. We introduce index files in order to map the full state of a file system to a single data structure. We divide the system into swarms of flexible size so that accessing data and monitoring is easy. Files are handled differently depending of their size in order to increase the system performance and reduce resource requirements. All files and their metadata are encrypted before being uploaded, so that data confidentiality is ensured. We further use state-of-the-art technologies in order to design a tracker that is scalable, fault-tolerant, and is replaceable in case it entirely leaves the system. The load of the tracker only depends on the number of participants, not on the amount of data stored in the network. The system allows configuring a time span within which a user needs to recover his data in case of local data loss. We analyze this approach by using real world connectivity traces of residential gateways and show that it results in low resource demands. Together with simulations on these traces, we underline the feasibility of our service.
