【作者】 计莉；

【导师】 阳春华；

【作者基本信息】 中南大学 ， 控制理论与控制工程， 2003， 硕士

【摘要】 实时系统越来越受到关注，正成为研究的热门领域，在国防、航空航天、自动控制等方面应用极为广泛。实时系统不仅要保证逻辑的正确性，而且要在确定的时间内提供正确的结果，否则就会导致整个系统失败，甚至引起灾难性后果。由于实时系统对时间特殊甚至是苛刻的要求，使得系统调度和系统容错成为该领域最重要的研究内容之一。本文以经典实时系统容错调度算法为理论依据，设计了基于不精确计算模型的启发式容错调度算法，有效改善实时系统的容错性能。其主要研究工作体现在以下几个方面： 考虑到经典实时容错调度算法无法直接处理系统超负荷(过载)的情况，设计了基于不精确计算模型(Imprecise Computation Model，ICM)的高效实时容错调度算法(Maximum Crucial First，MCF)。ICM模型为系统超负荷任务调度提供了一个比较灵活的框架，通过适当降低任务的计算精度来换取执行时间，使任务能在时间约束之内得出基本可用的结果，任务所获服务时间越长，其结果的精度越高。在MCF调度算法中利用单调速率调度算法决定强制性任务的关键级别，利用最早时限优先调度算法和最短空闲时间优先调度算法进一步确定关键集合中的任务优先级。该算法充分结合了静／动态容错调度算法的优点，最大限度地利用处理机。 现实世界中的实时任务具有关键时间限制的特点，利用ICM模型，以最大回报率，最小响应时间，最小误差为目的，寻找划分强制性实时任务和可选择性实时任务的最佳调度点，设计了基于不精确计算模型ICM的Optimal-Point容错调度算法。算法保证实时任务顺利调度的同时，最大限度地满足系统对回报率、响应时间和误差上的需求。 随着计算机网络规模的增大和复杂性的增加，当网络中某个组件失效时，网络管理系统必须迅速找到故障并及时排除。本文研究了基于简单网络管理协议(Simple Network Management Protocol，SNMP)的不精确计算模型ICM的实时网络容错系统软件设计，满足数字图像在网络传输中的高质量要求。更多还原

【Abstract】 Real-time systems are becoming more and more concerned, which are widely used in manufacturing, space or avionics, telecommunication, and industrial automation systems. The correctness of real-time systems depend not only on the result of computations, but also on the time instants at which these results become available, otherwise it will cause the whole system to fail, even lead to catastrophic consequences. Fault-tolerant algorithms become important research, which should guarantees the completion of a scheduled task before deadline in the presence of failures. Based on the classical real-time fault-tolerant scheduling algorithm, new heuristic fault-tolerant scheduling algorithms are proposed, which improve the fault-tolerant performance of the real-time system effectively. The main research in this paper include the following several respects: ICM model for task scheduling offers one flexible frame, and quantifies the trade-off between result quality and computation time. The imprecise computation technique can prevent timing faults and achieve graceful degradation by giving the user an approximate result of acceptable quality whenever the systems cannot produce the exact result in time. The mandatory subtask is required for an acceptable result and must be computed to completion before the task deadline. The optional subtask refines the result. Classical static scheduling algorithm (Rate Monotonic algorithm, RM) RM does not support dynamic systems very well, and (Earliest Deadline First algorithm, EDF) EDF and (Least Laxity First algorithm, LLF) LLF dynamic scheduling algorithms can’t deal with transient overload in the systems. Maximum-critical-first (MCF) combines the advantage of the static algorithm and dynamic algorithms. The critical set and tasks priority is scheduled by the RM and EDF. LLF respectively. Simulations are described to evaluate the performance of this algorithm, and show that the processors can achieve high utilization.> Optimal-Point fault-tolerant scheduling algorithm is proposed based on the Imprecise Computation Model (ICM). Considering that the real-time task in the practical world has rigidly time restriction. Which can divide mandatory real-time tasks and optional real-time tasks. The algorithm guarantees real-time task to finish at the deadline, at the same time, meets the requirements of reward, response time and error.> With the increasing of computer applications and complexity, some components loseefficiency in the network. Network administrative system must locate the fault and get rid of it rapidly. The real-time video transmission problem is to transmit video with timing constraints. This paper discusses the network fault-tolerant software design, which is managed by Simple Network Management Protocol (SNMP). The system guarantees the high quality requirement of the video transmitting through the network, which trades the quality of the transmitted video with the amount of available time.更多还原