异步DVE系统中生命周期约束下的因果一致性控制方法研究
A Lifetime-limited Causal Order Control Method in Asynchronous DVE System
-
摘要: 分布式虚拟环境是模拟现实世界的虚拟空间,对因果一致性控制具有实时性要求,必须在事件生命周期结束前得到维护.然而,在大规模网络条件下,网络传输高延迟和动态性会导致部分事件不能及时到达,使已传到事件间因果关系无法在生命周期限制内有效传递.在现有方法中,部分方法基于所有事件一定能及时传到的假设,没有考虑生命周期对因果关系的制约;而另一部分方法虽然考虑了生命周期的约束,但其因果关系传递要求仿真时钟精确同步,且因果控制效率随系统规模的扩大而快速降低,限制了虚拟环境的普适性和实时性.提出了生命周期约束下的因果一致性控制方法LCO,突破了异步时钟间的时间值比较、多路径因果控制信息选择的终止条件、网络状况敏感的因果控制信息动态调节等关键技术,能够在事件无法及时传到时,仍可以根据已传到的事件计算出因果传递关系.实验证明,LCO既能维护生命周期内的因果一致性,又使因果控制信息量与系统规模无关,降低网络传输和计算开销.Abstract: Distributed virtual environment(DVE) is a computer-generated virtual space that simulates the real world. Therefore, in a DVE, causal order consistency is required to be preserved in real time, which means the causal events must be delivered within the lifetime of the result event. However, due to the network latency, part of causal events may not arrive the receiving node in time especially in large-scale DVE, and then the causality between the arrived causal events and the result event can not be maintained within the lifetime of it. In related work, some do not consider the causality with lifetime limitation based on the presumption that all events can arrive in time, while others require accurate synchronous simulation clock and their control overhead is closely-coupled with the system scale so that causal control efficiency becomes very low in large-scale DVE. In this paper, we propose a novel lifetime-limited causal order (LCO) control method that can compare asynchronous time of different nodes, conclude the ending condition of multi-path causal order control information selection and dynamically adapt the causal control information according to network latency variation. Thus even when part of causal events can’t arrive in time, the causality among arrived events can be preserved within lifetime limitation using causal control information selected by LCO. The experiment results demonstrate that LCO can effectively preserve causal order consistency within lifetime and the overhead of the causal control information is irrelevant with system scale.