On the characterization of avatars in Distributed Virtual Worlds

Abstract

Current Distributed Virtual Environment (DVE) systems run simulations based on a server-network architecture, where the population of avatars should be properly assigned to the servers in the DVE. This goal, called partitioning problem, is a hot research problem in the field of networked 3D real time graphics. Some approaches have been proposed for resolving this problem, all of them based on a very basic model which describes the behaviour of an avatar in a virtual world. This model estimates the workload that a given avatar adds to its server as an independent and static value. In order to design scalable and accurate partitioning schemes we propose a new characterization of the workload generated by the avatars in the system. In this model, the workload is computed as a function of two factors: avatar movement rate and factor of presence. Evaluation results show that this proposed model correlates with real systems properly, providing an accurate abstraction of avatars for load balancing purposes.