Refine
Document Type
- conference proceeding (article) (10)
- Article (4)
Has Fulltext
- no (14)
Is part of the Bibliography
- no (14)
Keywords
- reliability analysis (3)
- error probability (2)
- Erlang-distribution (1)
- Markov model (1)
- Model (1)
- Stochastic simulation (1)
- addition (1)
- channel model (1)
- coded processing (1)
- continuous-time Markov process (1)
Institute
Begutachtungsstatus
- peer-reviewed (1)
The requirements for safety-related software systems increases rapidly. To detect arbitrary hardware faults, there are applicable coding mechanism, that add redundancy to the software. In this way it is possible to replace conventional multi-channel hardware and so reduce costs. Arithmetic codes are one possibility of coded processing and are used in this approach. A further approach to increase fault tolerance is the multiple execution of certain critical parts of software. This kind of time redundancy is easily realized by the parallel processing in an operating system. Faults in the program flow can be monitored. No special compilers, that insert additional generated code into the existing program, are required. The usage of multi-core processors would further increase the performance of such multi-channel software systems. In this paper we present the approach of program flow monitoring combined with coded processing, which is encapsulated in a library of coded data types. The program flow monitoring is indirectly realized by means of an operating system.
In this paper we present simulation and model based approaches for evaluating and validating the temporal and safety relevant properties of software intensive safety-critical real-time embedded systems. A high level reliability model of a safe task execution is described by a continuous-time Markov process, enhanced by the modeling of execution times. It is shown that the behavior - regarding real-time and safety metrics - of this theoretical model can be transferred into an abstract system timing model, which then can be analyzed by a discrete event simulation approach. The verification of the discrete event simulation by Markov models offers the possibility of a holistic approach for reliability analysis combined with schedulability analysis of complex safety-critical multicore real-time systems by the discrete event simulation.
Error detecting and correcting codes are widely used in data transmission, storage systems and also for data processing. In logical circuits like arithmetic operations, arbitrary faults can cause errors in the result. However in safety critical applications, it is important to avoid those errors which would lead to system failures. Several approaches are known to protect the result of operations during software processing. In the same way like transmission systems, coded processing uses codes for fault detection. But in contrast to transmission systems, there is no adequate channel model available which makes it possible to evaluate the residue error probability of an arithmetic operation in an analytical way. This paper tries to close the gap of arithmetic error models by the development of a model for an ordinary addition in a computer system. Thus, the reliability of an addition's result can be analytically evaluated.
The data flow is a crucial part of software execution in recent applications. It depends on the concrete implementation of the realized algorithm and it influences the correctness of a result in case of hardware faults during the calculation. In logical circuits, like arithmetic operations in a processor system, arbitrary faults become a more tremendous aspect in future. With modern manufacturing processes, the probability of such faults will increase and the result of a software's data flow will be more vulnerable. This paper shows a principle evaluation method for the reliability of a software's data flow with arbitrary soft errors also with the concept of fault compensation. This evaluation is discussed by means of a simple example based on an addition.
We present a simulation-based approach to reliability analysis combined with a schedulability analysis of software intensive embedded real-time systems. In such a system not only does the software execution have to be hardened against soft errors, e.g., by means of coded processing or diverse execution, but also the real-time requirements have still to be met in the presence of such error to guarantee a safe operation of the system. For that reason, the influence regarding the real-time characteristics of a given sporadic error with a certain error rate is analyzed by means of a Monte Carlo simulation. Different safety design patterns are introduced and compared. Furthermore, the impact on the schedulability of an embedded system is discussed.
This paper presents the reliability evaluation of task execution during safe software processing. The standard method of duplication in a safety-critical application can also be applied for tasks in a software system. But in addition to this, there is also the possibility for coded task processing to increase the reliability and availability of software. The presented analysis covers the reliability analysis of a single, a duplicated and a coded task by the technique of continuous time Markov processes. Markov processes are often used for the reliability evaluation of safety-critical systems. We introduce a method to describe the execution time of tasks by means of enhanced Markov models and their solution by numerical methods.
In logical circuits, like arithmetic operations in a processor system, arbitrary faults become a more tremendous aspect in future. Modern manufacturing processes lead to less reliability and higher vulnerability of software execution to soft-errors. The correctness of certain results is important especially for safety–critical applications whose reliability depends on the fault-free execution of each single instruction and the dependencies between them. The more complex a software is the more unreliable the outcome is. But, there is a contrary effect. If the probability for multiple faults increases, there is also the chance that two faults compensate each other and the result is correct again. This paper presents the basic ideas for such a reliability evaluation of a software's data flow with arbitrary soft-errors and the effect of fault compensation. Further, this evaluation provides a possibility to compare different implementations of a data flow with respect to the reliability. This is shown by the comparison of two different error codes as alternatives for coded data processing.