Universität Duisburg-Essen
Verteilte Systeme
Design and prototypical development of a mechanism for delay-minimization of interacting participants of a distributed virtual environment
Prüfer Prof. Dr.-Ing. Torben Weis
Betreuer Dipl.-Inform. Sebastian Schuster
Bearbeiter Matthias Helling
Dauer 6 Monate

The peers@play project is a joint project of the distributed systems research group together with the universities of Mannheim and Hannover. Its goal is the creation of an interactive three-dimensional virtual world model. Such a model can be useful in virtual societies. They have gained much attention in the last years, in the World Wide Web (e.g. Web 2.0) as well as in 3D applications (e.g. Second Life). Technically, peers@play should be realized fully distri-buted based on a peer-to-peer architecture. We expect a high degree of heterogeneity of devices and communication networks, so users should be able to access our platform using any device to participate in the virtual society. Peers@play is aimed at having thousands of users take part in the same virtual world. To actually deliver an enjoyable experience, consistency must be ensured and the world must react fast to user actions. This becomes a huge challenge in a peer-to-peer network consisting of thousands of nodes with varying latency and bandwidth. To actually find out, whether the peers@play approach is feasible, several scenarios have to be tested. However, running real world tests with networks as large as planned for peers@play is not feasible. The only possible way is to run simulations. In order to simulate thousands of nodes running the peers@play software, a distributed simulation splitting the simulation workload between multiple hosts is needed. There are a few network simulators available. However, it is unclear whether they meet the requirements of peers@play with regards to network size and simulation speed. Finally, the simulator must be compatible with .Net, the platform peers@play is built upon. So far, no simulator implemented with .Net is available.

Current massively multiplayer online games rely on client/server architectures. However, such games are not easily scalable in the number of participants. Additionally, the acquisition and maintenance costs for servers are very high. Instead, massively multiplayer virtual environments can be built using peer-to-peer techniques. The detection of visible players is crucial for virtual environments. In client/server systems this task is simple because all players are managed centrally. In peer-to-peer systems the detection is more complicated, because players are managed in a distributed manner. This paper shows a mechanism to detect visible players in the virtual environment. Furthermore, the delay for message exchange between players is minimized to reduce short-term inconsistencies and increase interactivity.

 

© Universität Duisburg-Essen, Verteilte Systeme - Kontakt: webmaster@vs.uni-due.de

Impressum - Datenschutz