Scalable Interest Management for Multidimensional Routing Space Elvis S. Liu mcelvis@polyu.edu.hk Milo K. Yip mcmilo@polyu.edu.hk Gino Yu mcgino@polyu.edu.hk Multimedia Innovation Centre School of Design The Hong Kong Polytechnic University ABSTRACT Interest management is essential for scalable collaborative virtual environments (CVEs) which sought to reduce band- width consumption on the network. Most of the interest management systems such as Data Distribution Manage- ment (DDM) service of the High Level Architecture (HLA) concentrate on providing precise message filtering mecha- nisms. However, in doing so a second problem is introduced: the CPU cycle overheads of filtering process. If the cost in terms of computational resources of interest management itself is too high, it would be unsuitable for real time appli- cations such as multiplayer online games (MOGs) for which runtime performance is important. In this paper we present a scalable interest management algorithm which is suitable for HLA DDM. Our approach employs the collision detection method of I-COLLIDE for fast interest matching. Further- more, the algorithm has been implemented in our commer- cialized MOG middleware - Lucid Platform. Experimental evidence demonstrates that it works well in practice. Categories and Subject Descriptors C.2.4 [Computer-Communication Networks]: Distrib- uted System—Client/Server, Distributed Applications ; I.6.7 [Simulation and Modeling]: Simulation Support Systems Keywords Computer Games, Collaborative Virtual Environments, In- terest Management, Collision Detection, High Level Archi- tecture, Data Distribution Management 1. INTRODUCTION Collaborative virtual environments (CVEs) allow multi- ple users to interact in real-time even though those users are located differently around the world. Since the success of military simulation CVEs such as SIMNET [5] and DIS Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. VRST’05, November 7–9, 2005, Monterey, California, USA. Copyright 2005 ACM 1-58113-098-1/05/0011 ...$5.00. [1] in early 1990s, the attention of CVE development was no longer restricted to the research communities. Commercial applications such as multiplayer online games (MOGs) were also available for entertainment purpose. Recently, high bandwidth and low latency network environments allows the gaming industry to develop massively multiplayer on- line games (MMOGs) which support thousands of geograph- ically distributed players to participate in the same shared virtual world. As the CVE grows in terms of users and simulation entities, using scalable data replication schemes becomes one of the major requirements for CVE develop- ment. Since entity state broadcasting consumes significant network resources, message filtering by interest management schemes was introduced to address the problem. The basic idea of interest management is very simple. All participants should only receive data of the simulation en- tities they interest, rather than the entities in the whole shared virtual world. Various data filtering mechanisms have been proposed based on geographical distribution of the simulation entities, which are mainly divided into two categories: zone-based [11] and aura-based [9]. These ap- proaches, however, do not concern filtering aspects other than spatial information. For example, in a military simu- lation, a low rank soldier may not receive information that sends to his nearby generals. This can be done by additional filtering mechanisms provided by the High-Level Architec- ture (HLA) [3, 14] such as class-based filtering and value- based filtering through multidimensional routing space. The HLA message filtering mechanisms are very precise which ensure the participants receive the minimal set of en- tity states they interest. However, the more complex the filtering mechanisms, the more computational resources are required to process interest matching between subscribers and simulation entities. Consider a CVE with n update re- gions associated with the objects and m subscription regions specified by the participants, using brute force approach to match their interest is an O(nm) process. The situation becomes worse when the concept of multidimensional rout- ing space of HLA is introduced, the complexity of matching process in terms of number of comparisons becomes O(nmd) where d is the number of dimensions of the routing space. Additional process may also be required to match the up- date and subscription classes as well as the attribute sets for the class-based filtering mechanism. For real-time applica- tions such as MMOGs, the simulation entities may modify their update regions at every frame; clearly the brute force