一种新型可重构星载计算机体系架构研究

为满足未来航天任务对星载计算机的高性能、高集成化的要求,本文提出了一种基于MPSOC的可重构星载计算机体系结构,由标准处理器、应用处理器、可重构模块、专用ASIC、I/O接口构成,采用片上多处理器系统能够满足卫星不同的应用模式和数据处理要求,从而提高计算机的并行性和处理性能。同时星载计算机采用一种软硬件协同设计方法,可提升系统性能和可靠性,并缩短研制周期。     

容错的并行多重网格算法

高性能计算机上并行程序用到的结点越来越多,而在程序运行期间中发生结点失效的概率也随之增大.对于计算时间很长的程序,容忍结点失效的容错能力显得尤为重要.并行多重网格算法(MG)被广泛用于求解大型工程和物理问题中的偏微分方程组的数值解.为了实现MG算法的容错能力,提出了一种基于容错MPI的容错并行多重网格算法FT-MG.实验结果表明:FT-MG算法在引入少许开销的条件下实现了MG算法的容错能力.     

A Fault Tolerant Technique for FPGAs

In this paper we present a fault tolerant (FT) technique for field programmable gate arrays (FPGAs) that is based on incrementally reconfiguring circuits and applications that have been previously placed and routed. Our technique targets both logic faults and interconnect faults, and our algorithms can be applied to either static or run-time reconfigurable FPGAs. The algorithm for reconfiguring designs in the presence of logic faults uses a matching technique. The matching technique requires no preplaced, spare logic resources and is capable of handling groups of faults. Experimental results indicate there is little or no impact on circuit performance for low numbers of reconfigured logic blocks. For interconnect faults, we present a rip-up and reroute strategy. Our strategy is based on reading back the FPGA configuration memory, so no netlist is required for rerouting around faulty resources. Experimental results indicate high incremental routability for low numbers of interconnect faults. We also lay the foundation for applying our approach to yield enhancement.     

星载Clos网络的全分布式容错调度算法

针对星载交换结构受空间辐射影响造成的可靠性严重下降问题,该文提出了一种支持全分布式调度的三级Clos网络及其全分布式容错(Fully Distributed Fault Tolerant, FDFT)调度算法,以提高星载交换结构在交叉点故障下的容错能力。该Clos网络的中间级和输出级采用联合输入交叉点队列,以支持Clos网络和交换单元内部的全分布式调度。FDFT采用一种分布式故障检测算法获得交叉点故障信息。基于对交叉点故障影响范围的分析,FDFT在输入级采用一种容错信元分发算法,实现无故障路径的负载均衡。理论分析证明,当任一输入/输出级交换单元故障个数不超过(m-n)或所有中间级交换单元故障个数不超过(m-n)时,其中m, n分别为输入级交换单元输入、输出端口数,FDFT能够达到100%吞吐率。仿真结果进一步验证,故障随机发生情况下,FDFT能够抵抗比故障任意发生情况下更多的故障,且在不同的业务场景下具有良好的吞吐率和时延性能。     

The Fault-Tolerant Mechanism of the CIN01-A System

ＴｈｅＦａｕｌｔ┐ＴｏｌｅｒａｎｔＭｅｃｈａｎｉｓｍｏｆｔｈｅＣＩＮ０１┐ＡＳｙｓｔｅｍＭａｎｕｓｃｒｉｐｔｒｅｃｅｉｖｅｄＪａｎ．１６，１９９７．ＴｈｉｓｐａｐｅｒｉｓｓｐｏｎｓｏｒｅｄｂｙＣｈｉｎａ“８６３”ＨｉｇｈＳｃｉ．＆Ｔｅｃｈ．Ｄｅｖｅｌ...     

WiNoC中无线通信拥塞与故障感知的容错路由算法

在无线片上网络中,无线通信拥塞和故障对整个片上网络的通信效率具有严重影响.为此本文提出了一种针对无线通信拥塞和故障的容错路由算法,首先设计了无线通信拥塞和故障感知模型,该模型能够感知无线节点通信对的拥塞和故障信息,并对其编码发送给子网中的路由器;然后子网中的路由器根据接收到的无线节点通信对状态信息,判断数据包是否使用无线传输.实验表明,本文方案相较于对比对象能够在较小的额外面积、功耗开销下,保证较低的网络延迟和较高的网络吞吐率,并对无线节点通信对的永久性故障具有良好的容错能力.     

计算机硬件故障分类及诊断处理

计算机设备使用不当或意外受损,出现故障是难免的。要想快速准确的排除故障,除掌握必要的基本理论外,还需具备一定的检修方法和故障处理技巧。从故障分类入手,分析了计算机发生故障的原因,介绍了常用的维修方法和技巧。     

An Inter-cluster Communication based Energy Aware and Fault Tolerant Protocol for Wireless Sensor Networks

The ambient intelligence paradigm is built upon Ubiquitous Computing (UC), in which the computing devices are embedded in the environment with the purpose of enhancing the human experience at home, workplace/office, learning, health care etc. The UC applications aim at providing services to the users anywhere, anytime in an unobtrusive, seemingly invisible way. Wireless sensor networks (WSNs) have great potential for UC applications and are envisioned to revolutionize them. This paper presents a clustering routing protocol for event-driven, query-based and periodic WSNs. The protocol aims at optimizing energy dissipation in the network as well as providing network’s fault tolerance and connectivity. Message propagation is accomplished by using short distance transmissions by employing nearest neighbor nodes between neighboring clusters. Moreover, the algorithm proposes using an energy efficient approach by alternating the nodes responsible for inter-cluster communication inside one cluster. The algorithm also aims at even energy dissipation among the nodes in the network by alternating the possible routes to the Sink. This helps to balance the load on sensor nodes and increases the network lifetime, while avoiding congested links at the same time. We discuss the implementation of our protocol, present its proof of correctness as well as the performance evaluation through an extensive set of simulation experiments. This work is partially sponsored by Grants from the NSERC, Canada Research Chairs Program, ORNEC, the Ontario Distinguished Researcher Award and the EAR Award.     

用于低开销容错设计的存储部件可靠性评估研究

低开销容错技术是当前软错误研究领域的热点。为了对微处理器进行低开销容错保护,首先就需要对微处理器可靠性(即体系结构弱点因子AVF (Architectural Vulnerability Factor))进行准确评估。然而,现有的AVF评估工具的精确性和适用范围都受到不同程度的限制。该文以微处理器上的核心部件(即存储部件)作为研究对象,对AVF评估方法进行改进,提出了一种访存操作分析和指令分析相结合的AVF评估策略HAES (Hybrid AVF Evaluation Strategy)。该文将HAES融入到通用的模拟器中,实现了更精确和更通用的AVF评估框架。实验结果表明相比其它AVF评估工具,利用该文提出的评估框架得到的AVF平均降低22.6%。基于该评估框架计算得到的AVF更加精确地反映了不同应用程序运行时存储部件的可靠性,对设计人员对微处理器进行低开销的容错设计具有重要指导意义。     

A New Approach to Software-Implemented Fault Tolerance

A new approach for providing fault detection and correction capabilities by using software techniques only is described. The approach is suitable for developing safety-critical applications exploiting unhardened commercial-off-the-shelf processor-based architectures. Data and code duplications are exploited to detect and correct transient faults affecting the processor data segment, while control flow instruction duplication is used for detecting and correcting faults affecting the code segment. Results coming from extensive fault injection campaigns showed the effectiveness and the limitations of the method.     

可诊断容差模拟电路软故障的新故障字典法

基于支路屏蔽的原理,本文提出一种可以诊断容差模拟电路软故障的新故障字典法.文中讨论了该方法的诊断原理和字典的建立方法,给出了容差情况下的仿真诊断例子.     

一种新的前向神经网络部件冗余容错方法

多层前向神经网络(MLP)的容错性有两种主要的研究方法:改进算法和部件冗余.前一种方法需要耗用大量的学习时间,对大型网络是不适用的.Phatak曾提出了用后一种方法进行MLP的单故障容错的一种网络结构,但是冗余部件数庞大,尤其对于大型网络.本文提出了一种新的冗余体系结构,针对单隐层MLP的单故障容错问题.这种体系结构充分考虑了不同权值的不同重要度,解决了原体系结构的仅值瓶颈问题,可以显著减少冗余部件数,尤其对于大型网络,更具有优越性.     

计算机网络通信技术故障分析与处理研究

计算机网络通信设施持续发展,相应提升了通信设计水平.计算机网络通信工程建设为经济发展和生活发展提供了安全保障,因此必须高度重视通信工程建设,加大技术管理力度,才可以有效处理通信工程存在的问题,高效实施通信工程,从而推动通信建设发展.围绕计算机网络通信技术展开讨论,重点分析通信技术故障,提出科学化处理与控制措施,以供参考...     

接口通讯的软件容错设计分析

在弹载嵌入式软件设计中,需考虑各种故障模式并进行针对性软件容错设计。软件容错设计包含信息容错、时间容错和结构容错。对于实时性系统来说,接口通讯过程中受到干扰等外界因素会出现通讯数据异常的偶发性故障,针对该故障模式,在信息容错的基础上,进一步设计两种软件容错方案,并开展其风险分析。这两种软件容错设计方法的可行性和有效性均在工程实际应用中得到试验验证。     

一种基于时延补偿的星上分组调度算法

EDF策略具有有线调度算法中最大调度域和时延保障特性,根据分组时延要求确定服务优先级,是一种动态权重调度策略,通过将IQ-VOQ结构EDF策略引入无线调度并通过加入信道补偿机制和门限丢弃策略,得到一种新的算法IQCEDF,它避开公平排队固有的时延一带宽耦合问题,获得了比EDF更加出色的时延特性,更适于为卫星通信系统分组业务提供服务质量保证。     

A Unified Method for Analyzing Mission Reliability for Fault Tolerant Computer Systems

A reliability model is proposed and evaluated for a fault tolerant computer system which consists of multiple classes of modules and allows for degraded modes of performance. Each module of a given class has both an active and a passive hazard rate; constant hazard rates are assumed for active and dormant failures, and the given class may operate either in N Modular Redundancy (NMR: n + 1 out of 2n + 1 = N) or as a standby sparing system. The model allows for mission-phase changes at deterministic time points when the numbers of modules per class can be changed. The analysis proceeds by generalizing the notions of standby and NMR redundancy, which for N = 3 is TMR (Triple Modular Redundancy), into a concept called hybrid-degraded redundancy. The probabilistic evaluation of the unified redundancy concept is then developed to yield, for a given modular class, the joint distribution of success and the number of nonfailed modules from that class, at special times. With this information, a Markov chain analysis gives the reliability of an entire sequence of phases (mission profile).     

火控计算机系统可靠性的模糊故障树分析

传统故障树分析方法需要获取事件发生概率的精确值,忽略了数据的不完备性和模糊性.在火控计算机系统可靠性研究中提出模糊故障树分析方法.建立火控计算机系统的故障树,采用三角模糊数描述基本事件发生概率,计算得到顶事件发生概率的置信区间,并采用中值法计算基本事件的模糊重要度.结果更加符合实际情况,表明该方法用于可靠性分析合理可行.     

面向星载数据管理系统的SpaceWire总线仿真系统开发

本文介绍并分析了日本的X射线天文卫星ASTRO-H中SpaceWire的应用模式,然后结合我国星载数据管理系统建立了SpaceWire总线应用的仿真系统,模拟星载管理单元与设备以及设备与设备间的数据通信,实现了对系统综合数据业务的统一管理,对在轨数据系统实施自主管理和冗余容错管理奠定了技术基础.     

Nucleus Plus容错功能扩充研究

Nucleus Plus是为实时嵌入式应用而设计的一个抢先式多任务操作系统，其95％的代码使用ANSIC写成，非常便于移植并能够支持大多数类型的处理器。本文根据容错管理软件的设计思想，提出了一种Nucleus Plus实时操作系统容错功能扩充方法，并在某航天三模冗余计算机平台上进行了验证。     

On-Line Fault Monitoring

Sequoia's fault-tolerant computers were designed subject to some rather rigid constraints: No single hardware malfunction can generate an undetected error; an integrated circuit is a black box that can fail in arbitrary ways, affecting an arbitrary subset of input and output signals; faults can be transient or intermittent with arbitrary durations and repetition intervals. Moreover, the incremental hardware to be used to achieve these goals was to be kept to a minimum. The resulting computers do, to a very large extent, satisfy these constraints. To achieve this, a combination of fault-monitoring techniques was used, including: Bit and nibble error-correcting and error-detecting codes; byte parity codes with orthogonal partitioning; cyclic-residue codes on I/O data transfers; codes designed to protect against address counter overruns on I/O transfers; lossless control-signal compactors. The nature and rationale for these various fault monitors is described as well as the analytical and testing techniques used to estimate the resulting coverage.