Empirical Evaluation of the Efficiency of Spatial Subdivision Schemes and Load Balancing Strategies for Networked Games

Abstract

In this paper, we present an empirical evaluation of four spatial subdivision and three load balancing schemes that may be used for large-scale single-shard virtual environments, such as 3D networked Games and Virtual Communities. ALVIC-NG is an architecture that targets these environments and supports dynamic resource allocation. The scalability of an application based on ALVIC-NG and networked virtual environments in general is (among other factors) dependent on two metrics: the efficiency of the spatial subdivision scheme and the ability for the load balancing strategies to maintain a clustered assignment of regions to server instances. Both are discussed and evaluated in this paper through two modes of simulation. An offline version is used to decide on the most appropriate spatial subdivision scheme by creating a single-instance application that focuses on this specific metric. The online version implements the proposed strategies in ALVIC-NG and is used to both confirm the findings of the offline system and to determine the efficiency of the load balancing strategies by making use of several server instances and so-called bot servers to re-create the conditions of real-life deployments.