Distributed Personal Cloud Storage without Third Parties

Abstract

We present the “It's My Data file system” (IMDfs), a personal, version-tracking, distributed file system that allows users to take custody of their cross-device data management needs, without depending on cloud storage providers. Such a system may be desirable to privacy-conscious users, who worry about the increasing number of cloud storage incidents involving data leakage and corruption, or simply want to synchronize their devices without depending on third parties. IMDfs provides a global, unified view of a user's distributed data collection from any device, user-controlled replication of data across devices, transparent remote access, and support for non-restrictive conflict management. IMDfs achieves all this without sacrificing fault-tolerance or performance and with modest storage overhead. We present the design, implementation, and detailed performance evaluation of IMDfs using benchmarks and traces from real workloads, and we demonstrate that: 1) IMDfs, as a local file system, has reasonable overhead compared with a FUSE loopback file system and Wayback, a local versioning file system, and 2) IMDfs, as a network file system, outperforms NFS and Ori, a state-of-the-art file system for cross-device data management.