Mobile agent based framework large-scale collaborative virtual environment

Abstract

Collaborative Virtual Environment (CVE) provides an online shared virtual world for the geographically dispersed people to interact with each other. However, the scalability of the existing CVE systems is limited due to the constraints in processing power and network bandwidth of each participating host. The research for this thesis is motivated to improve the scalability of CVE within a heterogeneous computing and networking environment from the data communication perspective. It is achieved through a Mobile Agent based framework for large-scale CVE (MACVE). In MACVE, the large-scale CVE (LCVE) software system is decomposed into a group of mobile agents. Each mobile agent is responsible for an independent task to provide services to exchange or process the data flows in LCVE. The mobile agents do not bond with any fixed hosts and they can migrate or clone dynamically at any suitable participating host which includes traditional server hosts and qualified user hosts. By agent migration, LCVE tasks will not converge at a single host; by remote agent cloning, a single task will be shared by multiple hosts. The mutual independence of services and hosts provide large freedoms for LCVE system to utilize the system resource efficiently and avoid the potential bottleneck. The ability for qualified user hosts to take over the mobile agents makes the system’s scalability less confined by the number of traditional servers. Thus, the system data flows and workloads can be pervasively distributed. Moreover, MACVE provides adaptive consistency control to dynamically decide the consistency model and data communication architecture for different object state data depending on run-time activities and consistency requirements in each particular part of the LCVE, which will further enhance LCVE system scalability. This thesis contributes a mobile agent framework for large scale CVE (MACVE); a specially designed mobile agent system for MACVE which includes Mobile Agent Environment, MACVE APIs, and MACVE Protocol; a prototype CVE system with eleven basic types of agents and a web-based user application. Experiments with the prototype system have demonstrated the feasibility of MACVE. It also has shown the advantages of the proposed MACVE for improving the scalability of the CVE applications.