Communication Architecture and Interest Management for Large-Scale Virtual Environments
Original Publication Date: 2004-May-07
Included in the Prior Art Database: 2004-May-07
This article describes a mechanism, a.k.a interest management system, for partitioning large virtual environments into multiple regions. The mechanism allows to maintain these regions in a distributed fashion with low maintenance cost, which improves scalability comared to previous systems.
Communication Architecture and Interest Management for Large -Scale Virtual Environments
In large-scale distributed simulations, large multiuser virtual environments and massive multiplayer online games, thousands of entities interact over a network. A key concern is the scalability of the system. Scalability is increased by using an interest management system that aims to deliver packets only to the subset of entities for which the packets are relevant. Existing interest management systems divide the simulation space into cells. Each cell is associated with a multicast group. Entities exchange information only with other entities belonging to the same cell, i.e. they subscribe to the multicast group associated with the cell. This approach has been described in the following papers:
Macedonia et al.: "Exploiting Reality With Multicast Groups," IEEE Computer graphics and Applications, Vol. 15, pp. 38-45
K. Morse: "Interest management in large-scale distributed simulations." Tech report 96-27, Dept. of Information and Computer Science, University of California, Irvine,
John W. Barrus, Richard C. Waters, and David B. Anderson. Locales: Supporting Large Multiuser Virtual Environments. IEEE Computer Graphics and Applications, 16(6):50--57, November 1996.
The cell-based solution is static and does not adapt itself to the distribution of entities within the system. For example, if a large number of entities are located in a single cell, then the scalability problems still occur. This can be prevented by creating very small cells, which in turn increases the overhead significantly, since for each cell a multicast group has to be maintained. Using small cells, the majority of the cells won't be used. A solution to this problem has been proposed in:
E. Lety and T. Turletti: "Issues in Designing a Communication Architecture for Large-Scale Virtual Environments", "International Workshop on Networked Group Communication", pp 54--71, 1999
This solution dynamically partitions the simulation space into cells of different sizes such that the entity density never exceeds a threshold. Although this solution prevents overload situations, it can not solve the following limitations which are inherent in all cell-based approaches:
cell-based approaches are entirely based on spatial information (i.e. the location of entities within the virtual environment). Other attributes such as entity type or membership in a group are not considered. cell-based approaches require a handover management when an entity leaves a cell
and enters another cell, or the cells are resized. This is a critical operation that requires the entity to synchronize with the new cell. In addition, joining a multicast group involves a "join latency" during which no information can be exchanged. As a consequence, a handover always induces a pause. Entities assigned to the same cell not necessarily have overlapping interests. Consequently, cell-base...