【作者】 张进；

【导师】 王煦法； 李金龙；

【作者基本信息】 中国科学技术大学 ， 计算机应用技术， 2008， 博士

【摘要】 随着计算机网络技术和虚拟环境技术的发展,分布式虚拟环境的理论及应用研究已成为网络技术研究的一个热点。由于P2P技术在实时性、可靠性、负载均衡和可扩展性等方面的优势,使用P2P技术代替传统的Client/Server技术来构建分布式虚拟环境已经成为一种趋势。在基于P2P的分布式虚拟环境系统中,虚拟空间中的对象的状态由节点局部计算和分发。由于缺少中央服务器的全局状态管理,因此基于P2P的分布式虚拟环境中还存在着时空一致性问题,因此基于P2P的分布式虚拟环境中还存在着时空一致性问题。在空间上,时空一致性要求节点发现的交互节点要与事实一致;相同Avatar对象的状态在不同节点中的不同备份要一致;节点接收到的事件和状态消息也要与事实一致。在时间上,时空一致性要求系统能实时地保证这三类一致性。因此,时空一致性要求系统能解决三个子问题:节点发现问题,节点同步交互问题以及节点防欺骗问题。本文针对基于P2P的分布式虚拟环境中的时空一致性问题,对节点之间的协同行为进行了深入的研究,并提出了相应的解决办法。1.为了提高基于P2P的分布式虚拟环境中节点发现的拓扑一致性、实时性以及可扩展性,本文通过借鉴生物内分泌系统中的荷尔蒙机制,提出了基于荷尔蒙的节点发现模型HP2PON,该模型利用荷尔蒙消息在网络中的扩散来实现节点发现。另外,本文还提出了一种基于P2P的分布式虚拟环境的节点发现分析框架。该框架通过保证网络拓扑的全局连通性、AOI邻居封闭性、邻居封闭性、薄弱区周期关注性来保证网络拓扑的局部感知性。基于该框架,本文提出了一种具有高拓扑一致性的完全分布式的节点发现模型SimWorld。实验和分析证明HP2PON模型和SimWorld模型的拓扑一致性都接近100%。此外,HP2PON模型还有很好的实时性,而SimWorld模型则具有很好的可扩展性。2.为了提高基于P2P的分布式虚拟环境中节点同步交互的实时性、负载均衡和可扩展性,本文结合P2P技术和Client/Server技术各自的优点,提出了单点计算同步交互模型。该模型把分布式虚拟环境中的全局状态的计算和分发均匀分配给系统中的所有节点,并且每个对象的状态只有一个计算源。节点在用户设定的“轮”长下进行交互,速度慢的节点将被踢出系统。实验和分析证明,单点计算同步交可模型中的通讯量约为C/S结构的2倍,每轮交互只需要2个应用层hop,网络中的带宽消耗和计算任务被均匀分配到各个节点中。3.为了提高基于P2P的分布式虚拟环境中节点防欺骗交互的实时性、负载均衡、可扩展性和防欺骗性,本文通过借鉴科研工作中的论文盲审机制,提出了单点代理计算防欺骗模型SPC。在SPC模型中,每个节点控制的Avatar对象状态的计算和分发部是由一个对应的代理计算节点管理。由于代理计算节点随机分配,因此Avatar对象的代理计算节点与该Avatar对象的控制节点之间,以及与该Avatar对象交互的Avatar对象的控制节点之间没有利益相关系性,从而有效地防止了欺骗行为。实验和分析证明,单点代理计算防欺骗模型能够解决基于P2P分布式虚拟环境中目前涉及的各类协议层欺骗,而消耗的带宽只是C/S模型的3～4倍,带宽消耗和计算任务被均匀地分配到系统的所有节点中,每轮交互只需要2个应用层hop。更多还原

【Abstract】 With the development of the computer networks and virtual environment technology, research on the theory and application of distributed virtual environment (DVE) has become a hotspot of computer networks technology. Owing to the advantages of peer-to-peer (P2P) techonology in real-time property, security, scalability and load-balancing, it’s a trend to use P2P technology instead of the traditional client/server (C/S) architecture to conctruct DVE system.The states of Avatar objects are computed and distributed by local nodes in P2P- based DVE system. Spatio-temporal consistency is still a problem to be solved in such system due to the lack of global administration services of central server. Spatially, three requirements as follows should be satisfied. First, the interactive node discovered by each node should keep consistent with the truth. Second, different replicas of the same avatar object’s state should keep consistency with each other. Three, the event message and the state message received by each node should keep consistent with the truth. Temporally, the above three requirements should be satisfied real-timely. Therefore, spatio-temporal consistency could be divided into three subproblems which are the node discovery, synchronization of interaction among nodes and cheat-proofing of nodes’ interaction.Aiming at the spatio-temporal consistency of P2P-based distributed virtual environment, further research is done concerning the cooperative behaviors among nodes. Furthermore, corresponding measures are put forward to solve some problems in point.1. In order to improve the topological consistency, real-time property and scalability of node discovery in P2P-based distributed virtual environment, a hormone-based node discovery model named HP2PON is proposed by referring to the hormone mechanism in biological incretion system. According to the model, node discovery is actualized by the diffusion of hormone message in the network. Meanwhile, an analysis framework of node discovery in P2P-based DVE system is put forward. According to this framework, local awareness of the network can be reenforced by guaranteeing the global connectivity, neighbors’ encapsulation property, AOI neighbors’ encapsulation property and the periodical checking to the weak areas. On the basis of the framework, a fully distributed node discovery model with high topological consistency named SimWorld is proposed. Experiments and analysis proved that both HP2PON and Simworld model has approximate 100% topological consistency.2. In order to improve the real-time property, load-balancing and scalability ofthe synchronize mechanism in the P2P-based DVE, this paper proposes single point computing synchronize model. The task of global state’s computing is assigned uniformly to all the nodes in the system, and for each object there is only one source to compute its state. The interval of round is set by the user, and the slow node will be booted out of the system. The experiment and analysis prove that the traffic in single point computing synchronize model is approximately twice as much as the traffic in C/S, only two application level hops are needed in each round, at the same time, the bandwidth consumption and the computing task are distributed in all of the nodes uniformly.3. In order to improve the real-time property, load-balancing and scalability of cheat-proofing mechanism in the P2P-based DVE, this paper proposes single proxy computing model. The computing and distributing tasks of each avatar’s state are managed by a corresponding proxy node in the model. And due to the random assignment of proxy node, there is no profit relation between the proxy node and the controlling node of the same avatar object, and there is also no profit relation between the proxy node of avatar a and the nodes which control the avatars interacting with avatar a. so the model effectively proof the cheatings. The experiment and analysis prove that single point computing synchronize model can successfully proof several protocol cheatings, while the traffic in single proxy computing model is approximate three to four times as much as the traffic in C/S, and only two application level hops are needed in each round.更多还原