Collaborative Virtual Environments

Collaborative Virtual Environments, Networked Virtual Reality, Distributed Graphics

Virtual environments as collaboration tools and a way of sharing virtual scene among the users grow on importance. They are used for two main reasons: collaboration of more users working on common (virtual) data model - scene, and distribution of workload among the computers resulting in higher performance in data processing. Collaboration of more users is especially important today. People want to meet others without need of travelling, to discuss something without need to get up from the computer. Often people have some virtual data they need to share and discuss, for example a team of doctors discussing problem of broken skull of some patient whose virtual skull model they have on their computers. They want to put marks on model and want everybody to see everything that others have done with the model. Most of recent approaches are based on scene replication among the computers and serialization of events in the system as a convenient way of achieving scene consistency. The approach presented in this paper treats the scene as a replicated database and all updates as transactions sent to the system. Although the transactions are serialized for consistency reasons, peer-to-peer approach is used eliminating bottleneck of centralized solutions. The processing of transactions is the crucial problem of the whole system. Thousands of transactions in each second have to be potentially executed and they must be executed as soon as possible because CVE application runs usually in real-time and an user can not wait several seconds for the transaction to be finished as can be tolerated in classical databases. Therefore the traditional transaction processing model has to be modified for our needs. Especially, the client creating the transaction has to "precompute" the read and write sets of the transaction. The "precomputation" makes the transaction execution much more straightforward; specifically, it is very easy to detect concurrency violations among the transactions. Such approaches promiss high scalability, good performance and gives the user convenient way of keeping the scene consistent. The poster will describe the proposed method in more detail and present the benefits of the method.