计算系统软件抗衰重启技术研究

细粒度的软件抗衰，即部分的重启,它可以进一步降低抗衰成本,提高系统可用性.制定细粒度的软件抗衰策略的一个重要前提是确定可以安全重启的对象,并建立重启树.本文根据典型的软件体系结构和模块间的控制、调用及数据访问关系,研究了模块间的耦合程度和重启相关性之间的关系,给出了重启相关性和相关度的判定方法,确定了各模块的可达集和重启群,将模块的重启群作为抗衰策略的重启对象,并依据模块对系统资源的损耗程度,建立了重启树,制定了软件抗衰策略,建立了抗衰策略实施过程的有限自动机模型.

Research on Restart Technology for Software Rejuvenation of Computing System

Fine-grained software rejuvenation, that is partial restart technique, can further reduce the rejuvenation cost and improve the system availability. Before formulating the fine-gained software rejuvenation policy, the objects that can restart safely must be determined and the restart tree of computing system must be built. Therefore, based on the typical software architecture and the control, invoke and data access connection among modules, this paper studies the relation between the module-module coupling and the restart dependence, defines the restart dependence and its degree, and obtains the restart reachable set and restart group of each module. Taking the restart groups as the restart objects of rejuvenation policy, the restart tree can be built according to the wastage degree of system resource incurred by modules. Finally, the software rejuvenation policy is determined, and the implementation process of the policy is modelled by finite-state automaton.