| Abstract |
The commercialization of the Internet, the fast growth of its population and our society's increasing reliance on it as a means of information exchange, triggered the design and implementation of big, fast and reliable web services such as mail services, picture galleries and file hosting servers. Most of these services are independent of each other, usable only through their custom web interfaces and viewed as islands of functionality from their various users.
In this thesis we investigate an alternate view of these services. We utilize them as connected components of a new storage service instead of isolated entities. We design a system that acts as a storage parasite of the various free online infrastructures and aggregates their functionality into a new, free storage service. We implement an abstract API based on generic put and get operations which transcend each service's specific mechanisms and web interfaces. We use our framework to create a user-level file system, ParasiticFS, which can mount several distinct web services under a single unified namespace. All files shown locally in the ParasiticFS are in reality stored in various online services taking advantage of each service's unique functionality. ParasiticFS provides zero-cost storage that is globally accessible and characterized by high reliability and availability. Moreover, our system promotes a set of unique advantages such as NFS-like semantics, disaster recovery mechanisms and instant recovery of deleted files.
The resulting system and its performance suggests that our storage parasite solution can be practically used as a free and efficient backup service, a media filesystem or an all-purpose network storage solution.
|
Parasitic storage systems : free, reliable and globally accessible Gigabytes
