PDES中乐观时间同步的时空损耗研究

并行离散事件仿真对复杂大规模动态系统的研究以及探索其长远的应用提供了便利，近年来日益成为研究的热点。然而时间同步管理是影响并行离散事件仿真系统高效运行的重要因素之一。乐观的同步是采用检测和回退机制，允许逻辑进程积极的处理本地事件。一旦出现同步错误则利用回退机制从错误中恢复到较早状态，然后再恢复执行。这一切都是通过基于检查点状态保存重建机制来实现的，因而状态保存及状态重建必然伴随着时间和空间的损耗。该文深入研究了在乐观同步机制下，仿真执行时间和内存空间的损耗与检查点间隔之间的关系，并通过推理计算给出了检查点间隔的最优取值范围。

Time-Space Overhead Based on Optimistic Time Synchronization in Parallel Discrete Event Simulation

Recently there has been a great deal of interest in parallel discrete event simulation(PDES).It facilitates the research of large-scale complex dynamic systems.However,time synchronization is one of the important factors that affect PDES running performance.Optimistic synchronization uses a detection and recovery approach.A logical process(LP) is allowed to aggressively process local events.When a LP detects the error of synchronization,it returns to an earlier state to recover from such errors and then it resumes execution.This is accomplished by periodic state checkpoint and the desired state reconstruction from a checkpoint.Hence,there is the time overhead associated with the reconstruction of a state and the space overhead correlative with the checkpoint of a state.This paper emphatically discusses the relation between the checkpoint frequency and the time-space consumption.Finally,the scope of the optimal checkpoint interval is also presented by reasoning and calculation.