移动计算环境下检查点存储策略的研究

发布时间：2018-05-26 18:24

本文选题：移动计算 + 检查点技术　；参考：《哈尔滨工程大学》2012年硕士论文

【摘要】：与传统的分布式计算相比，移动计算系统具有一些固有特性，如移动主机存储空间小、能量低，无线网络带宽低以及无线连接易受干扰等。移动计算系统因为这些固有特性容易出现故障，如果没有针对移动计算故障的有效处理机制，将影响到移动计算系统正常运行。容错技术是保障移动计算系统正常运行的有效手段，而检查点技术又是一种常用的容错技术。在检查点技术的应用中涉及检查点数据的存储问题，检查点存储策略是否合理直接影响着移动计算系统的性能。本文将移动主机进程完成一个标准计算任务的平均时间作为衡量移动计算系统性能的指标，时间越少则表示系统的性能越好。经过分析，移动主机完成一个计算任务平均时间随着平均额外时间开销的增大而增大。系统往往在移动主机握手迁移时对检查点数据的存储位置进行调整。目前主要有三种检查点存储策略包括Eager策略、Lazy策略以及Trickle策略，但是它们在迁移检查点时带有盲目性，在移动主机进程完成一个标准计算任务过程中引入了较大的平均额外时间开销。本文分析比较了几种传统检查点存储策略的优点与缺点，本着继承各策略优点，同时摒弃各策略缺点的思路，提出了基于移动主机进程故障概率检查点存储策略。该策略将移动主机进程运行状态的稳定性作为是否进行检查点迁移的依据，避免了检查点迁移的盲目性，能够保证在所有的移动主机进程故障概率下都能引入较小的平均额外时间开销，从而降低移动主机进程完成一个标准计算任务的平均时间。本文给出了移动主机进程故障概率模型，用以计算移动主机发生握手迁移时，移动主机进程在下一个区域发生故障的概率，从而判定移动主机进程运行状态的稳定性。最后，，本文对各策略进行了理论上的比较分析，并通过实验验证了分析结果。结果显示基于移动主机进程故障概率检查点存储策略能够使移动主机进程以更小的平均时间完成一个标准计算任务，从而提升了移动计算系统的性能。
[Abstract]:Compared with the traditional distributed computing, the mobile computing system has some inherent characteristics, such as small storage space, low energy, low bandwidth of wireless network and vulnerable to interference of wireless connection. Mobile computing systems are prone to failure because of these inherent characteristics. If there is no effective mechanism to deal with mobile computing failures, it will affect the normal operation of mobile computing systems. Fault-tolerant technology is an effective means to ensure the normal operation of mobile computing system, and checkpoint technology is a commonly used fault-tolerant technology. The application of checkpoint technology involves the storage of checkpoint data. Whether the checkpoint storage strategy is reasonable or not directly affects the performance of mobile computing system. In this paper, the average time for a mobile host process to complete a standard computing task is taken as an index to measure the performance of a mobile computing system. The less the time is, the better the performance of the system is. After analysis, the average time for mobile host to complete a computing task increases with the increase of average extra time cost. The system often adjusts the storage location of checkpoint data when the mobile host handshake migrates. At present, there are three kinds of checkpointing storage strategies, including Eager policy, Lazy policy and Trickle policy, but they are blind when migrating checkpoint. In the process of mobile host process completing a standard computing task, the average extra time cost is introduced. This paper analyzes and compares the advantages and disadvantages of several traditional checkpointing storage strategies, inherits the advantages of each strategy, and at the same time abandons the ideas of the shortcomings of each strategy, and proposes a fault probability checkpoint storage strategy based on mobile host process. In this strategy, the stability of the running state of the mobile host process is taken as the basis of whether to carry out checkpoint migration, and the blindness of checkpoint migration is avoided. It can ensure that the average extra time cost is small under the failure probability of all mobile host processes, thus reducing the average time for mobile host processes to complete a standard computing task. In this paper, a fault probability model of mobile host process is presented, which is used to calculate the probability of failure of mobile host process in the next area when handshake migration occurs, so as to determine the stability of mobile host process running state. Finally, this paper makes a theoretical comparative analysis of the strategies, and verifies the results through experiments. The results show that the checkpoint storage strategy based on the fault probability of the mobile host process can make the mobile host process complete a standard computing task with a smaller average time, thus improving the performance of the mobile computing system.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP338

【参考文献】