穿戴计算机的内核级检查点优化策略研究

Optimization of kernel-level checkpoint mechanism for wearable computer systems

为提高穿戴计算机的可靠性和可用性,提出了适用于穿戴计算机操作系统内核级检查点机制的优化检查点间隔的策略——适应穿戴计算机瞬时故障发生规律的步进式检查点(SP-CKPT)策略和基于内存数据量监测的内存监控门限法(MMTC)。前者针对可穿戴计算机故障频率的特点,动态地步进式调整检查点间隔,使检查点位置'逼近'系统失效的位置,有效减少计算损失;后者通过对系统内存数据量的监测,采用门限法控制检查点设置的时机,使系统尽可能在用内存资源最少的时间点来进行检查点的设置,减少检查点信息的存储量。这两种策略分别从时间和空间的角度优化检查点间隔。仿真结果表明,这两种策略结合,能够提高系统可靠性和可用性,减少无故障运行和卷回恢复阶段的开销。

To improve the reliability and availability of wearable computer, the paper presents the strategies for optimization of kernel-level checkpoint mechanism for wearable computer systems the technologies of the step-checkpointing （SP- CKFF） and the memory monitoring and threshold-control（MMTC）. According to the characteristics of software or hard-ware environment-dependent system, SP-CKPT modulates the checkpoint intervals on-line. When failures occur frequently, checkpoint interval is set to the minimum length, then increased exponentially, in order to promptly normalize the checkpoint interval. MMTC adjusts the checkpointing interval by predicting the amount of data to be saved to implement the kernel-level memory-exclusion. SP-CKPT and MMTC respectively optimize the checkpoint interval from the aspect of time and space. The simulation results show the two schemes can effectively improve the performance of wearable computers and reduce the cost both in failure-free mn and rollback recovery.