Dissertation
Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Computer simulation

Research on Scalability of Collaborative Virtual Environment Systems

Author HuXiaoMei
Tutor CaiXiaoBin
School Northwestern Polytechnical University
Course Computer Science and Technology
Keywords Collaborative virtual environment Scalability Partitioning Load balancing Time-bound box
CLC TP391.9
Type PhD thesis
Year 2007
Downloads 189
Quotes 1
Download Dissertation

Collaborative Virtual Environment (CVE) is a sharing virtual space maintained by a group of computers interconnected through networks, used to support effective communications between the users to achieve coordination tasks. With the expansion of the scale of applications in distance education, collaborative design, joint military training and online multi-user game, improving the scalability becomes a hot topic of research in the field of CVE. The study of scalability of collaborative virtual environment systems aims that the software structure does not need to do major modifications in order to support the expansion of application scale, keeping the operating efficiency. Due to its importance both in theoretical research and practical application, scalability of collaborative virtual environment systems needs to be further studied. The research work and innovative contributions are as follows:1. A task-clustering based fix CVE partitioning algorithm is proposed. With this algorithm, the users which take part in the same task are assigned to the same regional server. The experimental results show that this algorithm gets the similar partitioning performance but takes much shorter execution time than that of ACS algorithm which was proposed by Morillo P. in Spain.2. The existing multi-server architecture has the defect of lacking the general management of CVE. In this paper, a hiberarchy based on tailoring tree model is proposed. The main server, as the control center, partitions the CVE according to the number and distribution of users in CVE by real time detection. A task-clustering based dynamic adaptive partitioning algorithm is proposed, which can adjust the partitioning number according to the number of users in the CVE. The experimental results show that the optimal partitioning performance and average executive efficiency of the task-clustering based fix partitioning algorithms are reached.3. Users are grouped according to their Area of Interest (AOI) in order to reduce the network traffic. A user grouping algorithm based on template matching is proposed to reduce the consumption of the network resources and computation resources. The experimental results prove that resource consumption is from 5% to 10% less than that of the cell-based grouping algorithm and the tracking-needless grouping algorithm.4. The active dynamic load balancing algorithm based on interest membership degree is proposed. Compared with the passive load balancing algorithm given by Lee K. in Korea, this algorithm has the similar ability of balancing computation load among the regional servers, but it gets less inter-server messages. System maximum response time is no more than 200ms under the active load balancing algorithms while over 450ms under passive load balancing algorithms. The experimental results prove that this algorithm meets the requirement of the real-time response of the system given by Park K. S. in Illinois University, U. S.5. Users send update messages in order to maintain the consistency of the CVE system, which increases the network traffic. A time-bound box based consistency maintenance algorithm is proposed. Compared with the accepted Dead Reckoning (DR) algorithm in CVE, this algorithm has the similar performance in reducing the network traffic; but entity collision error reduces 8%, even more when the network has a serious delay. The experimental results prove that this algorithm improves the consistency of the system.

Related Dissertations
More Dissertations