Research On Dynamic Active Interest Management Based On Differentiated Data Service In Distributed Virtual Environment

As one of the key technologies in solving communication scalability issues in Distributed Virtual Environment (DVE), Interest Management (IM) has been extensively studied. Currently, the interest management consists two mainstream mechanisms:the active router based active interest management and multi-server based partition mechanism. To provide better scalability for DVE, the active interest management aims to reduce the data transmission within the system, while the partition mechanism is to divide the whole workload to reduce workload and communication costs on a single serve. However, current studies have not taken the different processing requirements of differentiate services into consideration. Such approaches are weak in full use of resources and guaranteeing different services requirements. Therefore, it is an important aspect of scalability to introduce differentiated data services into interest management mechanisms to improve the efficiency of matching and forwarding data, and optimizing system resources. In addition, existing studies have not directly put the system performance into interest rules, hence still exists room for improvement in dynamic adjusting system performance. In this thesis, we study the two main interest management mechanisms in providing differentiated data services and dynamic adjustment based on the system performance, then give out two dynamic solutions of the interest management. The main work of this paper is as follows:(1) A deep research on current studies of existing communication scalability researches in distributed virtual environment systems. Moreover, compare and analyze the two mainstream interest management techniques which include active interest management and multi-server based partition, show their similarities and technology characteristics. Finally give out that there exits shortcomings of differentiated data service processing and dynamic system performance adjustment in current research of interest management in virtual environment.(2) Through a quantitative analysis of the importance of delay attribute in different service requirements in the active interest management, we choose delay as a benchmark to propose a differentiated data service-based dynamic active interest management. First, from the perspective of effectiveness of interactions, we design and implement the delay based dynamic access control to reduce invalid hosts joining in the system and redundant data transmission; then, propose a new interest rules including delay as a parameter, including dynamic interest subscription, matching and filtering, so as to reduce the invalid data transmission.(3) In the partition technique, we propose a service-based two-layered adaptive partition mechanism to provide differentiated services and dynamic adjustment of system performance in. Firstly, propose a two layered partition framework composed by global layer and local layer to reduce the cost of information maintenance and guarantee the correctness of partition; secondly, propose a service-based cost model and service-based migration strategy to deal with different services; thirdly, propose an adaptive quality function including hardware factors and application level factors as parameters to evaluate partitions. These three parts form the entire mechanism.(4) Based on the two designed mechanisms above, we implement two prototype systems in LAN, to finally bring the differentiated data processing and dynamic adjustment of system performance into the interest management technology and give out their analysis and simulation evaluations. The experimental results show that the system with the delay based dynamic active interest management mechanism is superior to system with current active interest management. In the service-base two-layered adaptive partition system, compared to avatars and static objects with scene information, dynamic objects with many services play more important impacts on system performance. The partition result using the service-based partition decision and workload migration is better than partitions based on avatars and scenes. Therefore, the two differentitated data processing based dynamic interest management mechanisms are feasible and effective and further provide a viable direction for providing diverse data services and dynamic control in the partition and the active interest management.