基于处理器核分配方案的众核处理器可靠性增强技术

发布时间：2018-04-20 04:17

本文选题：众核系统 + 单粒子效应　；参考：《上海交通大学》2014年硕士论文

【摘要】：随着航天器控制系统复杂度的提高，众核系统逐渐成为航天领域关注的技术方向之一。众核系统性能优越但对辐射的敏感度高。本文研究太空辐射条件下单粒子效应导致的失效问题，提升众核系统的可靠性。众核系统中，总线担当着通信与链接的重任，论文首先研究了如何进行总线参数的优化取值，使总线获得最高可靠性；论文然后研究了将众核系统中的空闲处理器核分配给各子系统以提高整个系统可靠性的分配策略。论文首先研究了众核系统总线的可靠性问题，在分析总线位宽、时钟频率等参数与故障率之间关系的基础上，建立用于评估总线可靠性的数学模型。结合对位宽、时钟频率和带宽等参数的约束条件，给出了在不同的约束条件组合下，使总线达到最高可靠性的参数取值。论文探讨了众核系统中冗余处理器核的优化分配方案，该方案基于可用度影响度优先‖的资源分配策略，选择对系统可用度影响最大的子系统优先进行分配。相比遍历所有冗余资源分配方案寻找最优解的方法，论文提出的分配策略运算复杂度低，能够实现冗余资源的实时动态调度，，提升众核处理器在空间环境下的生存能力。该方案包括子系统对整个系统可用度影响程度的快速计算方法、基于马尔可夫模型的系统总体可用度精确计算方法、对应冗余处理器核分配方案的系统状态转移图完整拓展方法。论文利用理论模型分析证明了系统可用度评估方法在准确度上的改进，最后利用仿真实验对所提的冗余处理器核分配方案的运算实时性和可用度提升量进行了验证。
[Abstract]:With the increasing complexity of spacecraft control system, multikernel system is becoming one of the technical directions in aerospace field. The multinuclear system is superior in performance but highly sensitive to radiation. In this paper, the failure problem caused by the order particle effect of space radiation conditions is studied to improve the reliability of the multinuclear system. In the multi-core system, bus plays an important role in communication and link. Firstly, this paper studies how to optimize the bus parameters to make the bus obtain the highest reliability. Then, the allocation strategy of allocating idle processor cores to each subsystem to improve the reliability of the system is studied. In this paper, the reliability of multi-core system bus is studied. Based on the analysis of the relationship between bus bit width, clock frequency and failure rate, a mathematical model is established to evaluate bus reliability. Combined with the constraints of the parameters such as bit width, clock frequency and bandwidth, the parameter values of maximum reliability of the bus are given under different constraint conditions. In this paper, the optimal allocation scheme of redundant processor cores in multi-core systems is discussed. Based on the resource allocation strategy based on the influence of availability, the sub-system that has the greatest impact on the availability of the system is selected for priority allocation. Compared with the method of traversing all redundant resource allocation schemes to find the optimal solution, the proposed allocation strategy has a low computational complexity and can achieve real-time dynamic scheduling of redundant resources and enhance the survivability of multi-core processors in spatial environment. The scheme includes a fast method for calculating the influence of subsystems on the system availability, and an accurate method for calculating the total availability of the system based on Markov model. A complete extension method of system state transition diagram corresponding to redundant processor core allocation scheme is presented. In this paper, the theoretical model analysis is used to prove the improvement of the accuracy of the system availability evaluation method. Finally, the simulation experiments are used to verify the real time operation and the availability enhancement of the proposed redundant processor core allocation scheme.
【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：V446;TP332

【参考文献】