Enabling Conflict-free Collaborations with Cloud Storage Services

Cloud storage services (e.g., Dropbox) have become pervasive in not only simple file sharing but also advanced collaborative file editing (collaboration for short). Using Dropbox for collaboration is much easier than SVN and Git, thus greatly facilitating common users. In practice, however, many Dropbox users are perplexed by unexpected collaboration conflicts, which severely impair their experiences. Through various benchmark experiments, we unveil the two root causes of collaboration conflicts: 1) Dropbox never locks an edited file during collaboration; 2) Dropbox only guarantees eventual data consistency among the collaborators, significantly aggravating the probability of conflicts. In this paper, we attempt to enable conflict-free collaborations with Dropbox-like cloud storage services. This attempt is empowered by three key findings and measures. First, although the end-to-end sync delay is unpredictable due to eventual consistency, we can always track the latest version of an edited file by actively resorting to the cloud via certain web APIs. Second, although all application-level data is encrypted in Dropbox, we can roughly deduce the sync status from traffic statistics. Third, applying a couple of useful mechanisms (e.g., distributed architecture and data lock) learned from Git, we can effectively and efficiently avoid collaboration conflicts-of course, this requires re-implementing Git mechanisms in cloud storage services with minimum overhead and user interference. Integrating above efforts, we build the ConflictReaper system capable of helping users automatically avoid almost all collaboration conflicts with affordable network and computation overhead.