摘要
操作系统故障根据传播特性可分为process-local和kernel-global 2类,分别造成进程局部数据和内核全局状态的错误。现有技术通过重启系统或故障进程实现对进程局部数据错误的恢复,但未考虑内核全局状态的不一致问题,不能保证对kernel-global类型故障的恢复效果。针对以上问题,提出了一种基于补偿回滚的故障自恢复技术。该技术通过监测内核全局方法调用,在进程局部数据被正确恢复的前提下,利用补偿操作对不一致的内核全局状态进行恢复,控制了故障的传播效应,减小了单点故障造成的影响。此外,该技术以内核模块的形式实现,不需要对目标操作系统进行修改,可便捷地实现功能扩展和移植。故障注入实验结果表明,在保证系统功能正常的前提下,该技术能对91.6%的故障进行有效恢复,且带来的系统负载较小。
Sections 1 through 4 of the full paper explain and evaluate a new fault recovery technique based on ker- nel motion compensation and process state roll-back. The core of our thinking and that of sections 1 through 4 con- sists of: (1) past research papers on operating system fault recovery mainly focus on the data loss caused by process-local faults and the global state inconsistency caused by kernel-global faults is neglected; we propose a new fault recovery technique based on kernel motion compensation and process state roll-back model; it can minimize the propagation of faults and ensure the consistency of global state ; this technique is implemented as loadable kernel module which makes it easy to expand functionality ; (2) section 2 presents the design of kernel motion compensa- tion and process state roll-back model ; (3) section 3 presents the implementation details of this technique in Linux operating system; (4) evaluation results presented in section 4 and their analysis show preliminarily the effective- ness of the proposed technique.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2015年第5期709-715,共7页
Journal of Northwestern Polytechnical University
基金
航天支撑技术基金(2013-HT-XGD(10))
陕西省科学技术研究发展计划项目(2014K05-25)
陕西省科学技术研究发展计划项目(2015GY035)
航空科学基金(20130753006)资助
关键词
操作系统
内核补偿
进程状态回滚
故障自恢复
adaptive algorithms, approximation algorithms, backstepping, conception design, cost functions,computer simulation, computer software, design, dynamic models, efficiency, embedded software,embedded systems, estimation, failure modes, fault detection, fault tolerance, global optimization,intelligent systems, mathematical models, models, motion compensation , real time control,reliability analysis, safety engineering, software reliability
fault recovery, kernel compensation, op-erating system, process state roll-back