High performance infrastructure for visually-intensive CSCW applications

We describe a scalable CSCW infrastructure designed to handle heavy-weight data sets, such as extremely large images and video. Scalability is achieved through exclusive use of reliable and unreliable multicast protocols. The infrastructure uses a replicated architecture rather than a centralized architecture, both to reduce latency and to improve responsiveness. Use of 1) reliable (multicast) transport of absolute, rather than relative, information sets, 2) time stamps, and 3) a last-in-wins policy provide coherency often lacking in replicated architectures. The infrastructure allows users to toggle between WYSIWIS and non-WYSIWIS modes. That, coupled with effective use of multicast groups, allows greatly improved responsiveness and performance for managing heavy-weight data.