Flockfs: A Moderated Collaboration System for Weakly Connected Workgroups

Collaborative file systems allow workgroups to share files. They can be used for a variety of purposes. For example, they can be used to share source code among a team of programmers, to produce a written report or to share multimedia contents that will later be assembled into a movie for a film course. Simultaneous modifications of the same file can either be addressed by requiring all other users to wait for updates to be completed or by using optimistic protocols that allow each user to independently operate on the various components of the project without cooperation with other group members. Optimistic protocols require a separate reconciliation of conflicting updates. Recent analysis of empirical wireless user availability data shows that users are migrating towards smaller session durations and larger duration between sessions. When users are away from the network, they can potentially create many more updates that will lead to a higher number of conflicts when they resume their network connection and the local file system attempts reconciliation with other nodes. I show that prior group collaboration systems that either used centralized or distributed approaches achieve poor performance. I show that maintaining a single shared copy is untenable for these weakly connected workgroups. Instead, I develop a moderated collaboration mechanism called flockfs that maintains multiple copies of the shared object. Moderation operations are similar to manual reconciliation operations in Coda and Ficus except that the moderator incorporates updates from all the other group members into their copy. Flockfs maintains one updateable copy of the shared content on each group member's node. It also hoards read-only copies of each of these updateable copies in any interested group member's node. For a group of size n, flockfs can potentially maintain one updateable copy (the author) and (n Ì¢è ' 1) read-only copies (from other users in the collaboration group). The various document versions will eventually converge into a single version through successive moderations. I show that flockfs avoids many of the problems of prior systems. Except for small groups, my distributed approach achieves similar performance as a server based system. My prototype exhibits acceptable file system performance and update propagation latency.