作　　者： ; ; (王鹏）; (赵亮）;

机构地区： 61226部队北京100079

出　　处： 《高技术通讯》 2012年第7期689-696,共8页

摘　　要： 针对多主体系统由其自治的软件主体构成带来的故障处理上的困难，研究了多主体系统的可靠性问题，并尝试从主体组层面出发对系统中的故障进行协调处理，提出了一个可靠关键子系统模型（RCSM）。RCSM包含三类主体，即中继主体、控制中心主体以及工作主体。此模型使用中继主体封装由工作主体组成的关键子系统，使子系统的拓扑结构对外界不可见，从而避免故障的大范围传播。在子系统内部采用基于任务分配的一致性检查点设置技术实现工作主体的故障恢复。对模型中的控制中心主体则采用了复制技术与团队协作相结合的手段来提高其可靠性。通过仿真实验对RCSM的故障恢复效率及其对系统正常运行效率的影响进行分析，说明了该模型的实用价值。 It was paid great attention that a multi-agent system is composed of autonomous software agents, which makes it more difficult to deal with the failures in the system than in a traditional distributed system, and the reliability problem of multi-agent systems was studied. This study tried to resolve the failures at the agent group level, and, as a result, a reliable critical subsystem model, called the RCSM for short, was introduced with the aim of improving the reliability of a multi-agent system. The RCSM consists of three types of agent： Relay Agent, Control Center A- gent, and Worker Agent. In RCSM, the Relay Agent encapsulates the critical subsystem, which is formed by Worker Agents, to avoid wide spread of failures by hiding the actual topology of critical subsystem. Within the sub- system RCSM adopts a consistent checkpoint mechanism based on task assignment to recover the failures in Worker Agent, and combines replication with teamwork to increase the robustness of Control Center Agent. The fault recov- ery efficiency of the RCSM and its influence on system load were analyzed by simulation experiments to illuminate the practicability of RCSM.

关 键 词： 多主体系统 可靠性模型 关键子系统 检查点 复制技术

领　　域： [自动化与计算机技术] [自动化与计算机技术]