多模冗余可重构计算机可靠性研究  

Study on Reliability of Multiple Modular Redundant Reconfigurable Computer

作　　者：李兴伟[1] 白博[1] 周军[1] 

机构地区：[1]西北工业大学,精确制导与控制研究所,西安710072

出　　处：《计算机测量与控制》2017年第7期309-312,316,共5页Computer Measurement &Control

基　　金：航天支撑技术基金(2015-HT-XGD);西北工业大学基础研究基金(JC201205)

摘　　要：为了以最小代价提高多模可重构计算机的可靠性,需要结合系统体积功耗等因素,对不同计算机体系结构的可靠性进行评估;首先,针对三模冗余、三热一冷以及四模热备这几种不同冗余可重构计算机体系结构,给出了他们的故障状态转移图;其次,运用马尔可夫模型理论,对这些计算机结构进行可靠性建模;第三,根据上述3种系统的可靠性模型,在考虑失效率和维修率两个参数对系统可靠性影响的情况下,对各结构的可靠性指标进行了评价;结果表明,三热一冷的可靠性大于三模冗余,而小于四模热备;但综合考虑系统的可靠度、体积、功耗等因素后,在这几种形式的表决系统中,三热一冷是最优的;同时,设计完善的三模可重构系统可靠度会比一个设计不合理的四模可重构系统要高。In order to improve the reliability of multiple modular reconfigurable computers with minimum cost, the reliability of different computer architectures need to be evaluated with system power and volume consideration. First, considering the redundant reeonfigurable computer architecture of triple modular redundancy （TMR）, quadruple modular redundant （QMR） and triple modular with one cold backup, the fault state transition diagrams of these architectures are proposed. Second, the reliability models of these computer systems are built with Markov model. Third, considering the influences of reliability with failure rate and maintenance ratio, the reliability of different structure are -evaluated based on the above three models. The comparison of three results indicates that, the reliability of triple modular with one cold is less than the reliability of QMR, while it is bigger than the reliability of TMR. But after considering the factors of reliability, system volume and power consumption, the triple modular with one cold is optimal. And the reliability of a perfectly designed TMR reconfigurable system will be higher than that of a bad designed QMR reconfigurable system.

关 键 词：多模冗余 马尔可夫模型 可重构计算机 

分 类 号：TP302.8[自动化与计算机技术—计算机系统结构]