| Latecomer and Crash Recovery Support in Fault Tolerant Groupware (2008) | |||||||||||||||
Abstract | |||||||||||||||
| In a real-time collaborative system, one user’s actions must be instantaneously propagated to all the other current participants. A session might start with some users, others might join in later, and still others might leave and join again. Unlike stateless collaborative applications such as videoconferencing and discussion groups, state-full collaborative applications must offer latecomer and crash recovery support. Such applications include shared electronic whiteboards, chat rooms, collaborative design, collaborative virtual worlds, and multiplayer games. There are two main research directions in this area: logging and playback of events, and exporting the application state to newcomers. Both have advantages and shortcomings. Event logging and playback imposes minimal requirements on applications and lets latecomers see how the existing participants reached the present state. However, it can require an excessive amount of storage, because sessions can last a long time. Bringing a latecomer up to date might also take a long time, for the same reason. The other technique, exporting the application state to each latecomer, has minimal storage requirements and takes relatively little time. However, the latecomer cannot see how the present situation developed. Also, the application must provide for exporting its data structures. The “Related Work ” sidebar points to other research in this area. Latecomer and crash recovery support should satisfy most or all of the following, sometimes contradictory, requirements: • resilience to individual site failures; • transparency to the user and application developer, being a general-purpose part of a collaboration infrastructure; | |||||||||||||||
Publication details | |||||||||||||||
| |||||||||||||||