实时系统的可靠性驱动任务调度机制研究

发布时间：2018-05-30 05:45

本文选题：实时嵌入式系统 + 任务调度　；参考：《华东师范大学》2017年博士论文

【摘要】：随着集成电路制造工艺的不断进步,当代微处理器芯片的尺寸已大幅缩小并且性能已大幅提高。这极大地促进了实时嵌入式系统在环境监测、空间探测、工业过程控制、飞行管理与控制、金融政务、汽车电子、智能电器以及医疗器械等诸多关键领域的广泛应用。但与此同时,微处理器芯片的晶体管数目和集成度也在持续增加,这使得微处理器受电磁干扰、瞬时电压扰动、高能粒子撞击等因素影响而产生瞬时故障的概率越来越高,从而导致系统的可靠性下降。然而,高可靠性是保证上述诸多应用成功运行的必然要求。因此,设计高可靠的实时嵌入式系统已成为必然趋势。可靠性驱动的实时任务调度机制是保证实时嵌入式系统高可靠性的重要手段,一直以来都是实时嵌入式系统领域的研究热点。该机制通过利用回卷恢复、冗余以及提频等技术来容忍已发生的瞬时故障或降低瞬时故障的发生率,从而提高系统的可靠性。但是,这些技术在提高可靠性的同时,也会带来处理器能耗增大、最大完工时间延长而导致吞吐量下降、温度升高而导致设备使用寿命缩短的问题。传统的可靠性驱动实时任务调度机制常常忽视了这些问题。因此,本文综合考虑上述问题,设计了新颖的可靠性驱动实时任务调度算法,在保证系统可靠性和实时性的前提下,优化系统的能耗、最大完工时间、以及设备使用寿命。具体来说,1.本文首先解决了可靠性和峰值温度约束下的能耗优化问题,提出了单处理器和多处理器系统中可靠性驱动的低能耗实时任务调度机制。在单处理器系统的研究工作中,本文提出了一个故障自适应因子并采用了检查点技术,以满足系统的可靠性约束;提出了一个能量有效因子来确定每个任务的工作频率,以最小化系统的能耗;提出了一种改进的热感知任务排序技术来降低处理器的温度,以满足系统的峰值温度约束。在多处理器系统的研究工作中,本文采用了所提的故障自适应因子和检查点技术以保证系统的可靠性,以及所提的热感知任务排序技术以保证系统的峰值温度约束;提出了一种动态能耗最优的任务分配方案,并且基于该方案,设计了相应的启发式算法,以实现多处理器系统中可靠性和峰值温度约束下的能耗优化。2.本文然后解决了可靠性和峰值温度约束下的最大完工时间优化问题,提出了多处理器系统中可靠性驱动的高吞吐量(即最大完工时间短)实时任务调度机制。在此研究工作中,本文基于混合整数线性规划方法,对多处理器系统最大完工时间的优化问题进行建模和求解;考虑到混合整数线性规划方法的不足,探索和分析了任务分配对最大完工时间、可靠性和峰值温度的影响;基于分析结果,提出了一种启发式的任务分配和调度算法,即利用调度长度均衡的任务分配、冗余技术、以及热感知的任务排序和调频技术,以实现可靠性和峰值温度约束下的最大完工时间优化。3.本文最后解决了可靠性和设备使用寿命的统一优化问题,提出了单处理器和多处理器系统中可靠性驱动的长使用寿命实时任务调度机制。在单处理器系统的研究工作中,本文提出了一种理论方法,用于计算瞬时故障对应的平均无故障时间;基于该计算方法,将单处理系统可靠性和设备使用寿命的统一优化问题转换为单处理器系统可用时间的优化问题,并设计了一个优化单处理器系统可用时间的框架。在多处理器系统的研究工作中,本文也是将多处理系统可靠性和设备使用寿命的统一优化问题转换为多处理器系统可用时间的优化问题,并设计了一个优化多处理器系统可用时间的方案。
[Abstract]:With the continuous progress of integrated circuit manufacturing technology, the size of modern microprocessor chip has been greatly reduced and its performance has been greatly improved. This greatly promotes the real-time embedded systems in environmental monitoring, space detection, industrial process control, flight management and control, financial affairs, automotive electronics, intelligent electrical appliances and medical devices and so on. But at the same time, the number and integration of the transistors on the microprocessor chip are also increasing. This makes the microprocessor affected by the influence of electromagnetic interference, instantaneous voltage disturbance, high energy particle impact and so on. The probability of the instantaneous failure is increasing, which leads to the reliability decline of the system. However, the reliability of the system is high. It is an inevitable requirement to ensure the successful operation of these applications. Therefore, it is an inevitable trend to design a highly reliable and real-time embedded system. The real-time task scheduling mechanism driven by reliability is an important means to ensure the high reliability of the real-time embedded system. It has always been a hot spot in the field of real-time embedded systems. Using the technology of rollback recovery, redundancy and frequency extraction to tolerate the instantaneous faults or reduce the occurrence of instantaneous faults, thus improving the reliability of the system. However, these technologies can also increase the reliability of the system, and also bring the increase in the energy consumption of the processor, the maximum completion time, resulting in the decline in throughput and the increase of temperature. The traditional reliability driven real-time task scheduling mechanism often ignores these problems. Therefore, a novel reliability driven real-time task scheduling algorithm is designed to optimize the energy consumption and maximum completion time of the system under the premise of ensuring the reliability and real-time performance of the system. In particular, 1. this paper first solves the problem of energy consumption optimization under the constraints of reliability and peak temperature, and proposes a low energy real-time task scheduling mechanism for reliability driven in single and multiprocessor systems. In the research work of a single processor system, a fault adaptive factor is proposed and used in this paper. A checkpoint technique is used to satisfy the reliability constraints of the system. An energy efficient factor is proposed to determine the working frequency of each task to minimize the energy consumption of the system. An improved thermal sensing task sequencing technique is proposed to reduce the temperature of the processor to meet the peak temperature constraints of the system. In this paper, the proposed fault adaptive factor and checkpoint technique are adopted to ensure the reliability of the system, and the proposed thermal sensing task sequencing technology is used to guarantee the peak temperature constraints of the system. A task allocation scheme is proposed for the optimal dynamic energy consumption, and a corresponding heuristic algorithm is designed based on the scheme. The optimization of energy consumption under the constraints of reliability and peak temperature in current multiprocessor systems.2. this paper then solves the problem of maximum completion time optimization under the constraints of reliability and peak temperature, and proposes a real-time task scheduling mechanism for reliability driven high throughput (i.e., the maximum completion time) in a multiprocessor system. In this paper, a hybrid integer linear programming method is used to model and solve the optimization problem of the maximum completion time of a multiprocessor system. Considering the shortage of mixed integer linear programming, the effect of task allocation on the maximum completion time, reliability and peak temperature is explored and analyzed. Based on the analysis results, a heuristic is proposed. Task allocation and scheduling algorithm, that is, using scheduling length balanced task allocation, redundancy technology, and thermal sensing task sequencing and frequency modulation technology to achieve maximum completion time optimization under the constraints of reliability and peak temperature.3. this paper finally solves the unified optimization problem of reliability and service life of equipment, and proposes a single processor and a single processor. The long service life real-time task scheduling mechanism in a multiprocessor system. In the research work of the single processor system, a theoretical method is proposed to calculate the average fault time corresponding to the instantaneous fault. Based on this method, the unified optimization problem of the reliability of single processing system and the service life of the equipment is presented. In the research work of the multiprocessor system, this paper also transforms the unified optimization problem of the reliability of multi processing system and the service life of the device to the optimization problem of the time of the multiprocessor system. A scheme to optimize the available time of multiprocessor systems is designed.
【学位授予单位】：华东师范大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP332

【相似文献】