Research On Supporting Techniques For High-Reliable Fault-Tolerant Real-Time Systems

Due to the development of real-time systems, more and more safety-critical applications become use real-time system as the kernel of the control system, and real-time systems become an affectable factor to determine the dependability of the safety-critical application. The dependable techniques of hardware of real-time have become mature enough, that makes software of real-time systems become the bottleneck of the system dependability. As a very important technique for system reliability, software fault tolerance has become a very hot topic.This paper analysis the factors that influence the system reliability, the causes of those factors, and the basic techniques of system reliability systematically. On the other hand, this paper introduces system and real-time scheduling in detail. Based on those researches, this paper researches fault-tolerant real-time system deeply. The author thinks that, the main obstacles of the development of fault-tolerant real-time system are:1. Fault tolerance is based on resource redundancy, which will reduce the resource utilization in the absence of faults. To solve this problem, the fault-tolerant real-time scheduling algorithms must be improved, to assure the fault-tolerant ability and increase the resource utilization at same time. Until now, there are lots of works that have been done in the realm of fault-tolerant real-time scheduling. Most of them are focused on theoretical research, and the higher resource utilization is at the cost of higher complex, that aren't practical in real world. So, the research and design of simple and practical scheduling is one of the most important topics of fault-tolerant real-time system.2. The construction and management of fault-tolerant real-time software increase the complexity of real-time system. Software fault-tolerance that realized in system level will increase the complexity and overhead of system software, reduce the running efficiency, and can't satisfy the requirement from special application. The realization in application level will make the burden on the application designers heavily, and make the control of the quality of software harder. So, how to keep the system level to be simple, common, and high efficiency, and at the same time, reduce the burden of application designer is another important research realm of fault-tolerant real-time system.Based on the analysis the main obstacle of fault-tolerant real-time system, this paper focus the view to the research of fault-tolerant real-time scheduling and architecture of fault-tolerant real-time system, and the main contributions are shown below: 1. Integrating imprecise computing and resource reclaim into fault-tolerantmreal-time scheduling, and present three new fault-tolerant real-time scheduling algorithms, IC-FT-RM, FTEDF, and RR-FTEDF. The object of those algorithms is to improve the resource utilization in the absence of faults, assure the fault-tolerant ability, and keep the algorithms simple enough. This paper proves the correctness of those algorithms, and validates them by simulating platform.2. Based on the principle of isolating the requirement of function and non-function, presenting to construct the fault-tolerant real-time software by fault-tolerant run-time library (programming middleware), to keep the system level to be simple, common, and high efficiency, and reduce the burden of application designer. This paper analysis the relation between system software, application software, and FTRTL. This paper gives an example to show how to realize FTRTL. Besides those works, this paper has done some works in the realm of theupgrade reliable of real-time system and safety kernel:1. Presenting a online upgrade model of real-time system, in the premise of assuring the function and performance, this model will help the real-time software upgrade reliable, to satisfy the change of environment and user requirement. This paper introduces the model and its composition in detail...