| Summary: | 碩士 === 逢甲大學 === 資訊工程所 === 94 === Massively Multiplayer Games (MMGs) is an interesting example of virtual environment (VE) applications on the Internet. In such applications, system communication performance becomes an important challenge. During game playing, a large amount of message will be produced and transmitted over Internet. Generally, most of these messages are unnecessary to players. Interest Management (IM) is a mechanism to filter out redundant messages and thus minimize communication traffics. With interest management mechanism, every player always defines an Area of Interest (AOI) to represent his interest. According to their AOIs, IM divides players into several groups, and associates each group with a multicast address. Unfortunately, multicast address is not an inexhaustible resource. Two well-known grouping techniques, entity-based grouping technique and cell-based grouping technique, have been proposed. Although entity-based grouping techniques will not result in any superfluous message, they need higher computation overheads, and thus not suitable for a large number of players. In contrast, cell-based grouping techniques are scalable to a large number of players, but they always generate much more superfluous messages. In this thesis, we propose a compromising cell-based grouping technique, named as consolidate-based grouping technique, used in MMGs to adaptively and effectively decrease the superfluous message rate while meeting the requirement on a limited number of available multicast addresses. In our approach, a VE is first partitioned into a number of cells, with a cell size smaller than that obtained from the traditional cell-based grouping strategy, to effectively reduce the superfluous messages. And then, an appropriate merging process is used to iteratively combine similar cells, such that the constraint on the number of available multicast groups can be met. According to simulation results, our approach can indeed reduce the superfluous message rate and achieve the desired requirement on the multicast addresses. Simulation results also show that, with threshold table look-up method, our approach works well around various densities in VE. Especially, our adaptive algorithm is more suitable for a slower speed VE.
|
|---|
