Reliability of Application Execution On A Processor with/without OS

In safety-critical applications, such as aerospace, automotive, medical, and industrial control systems, the occurrence of faults can have serious consequences, including loss of human life, property damage, and environmental damage. For this purpose, fault occurrence is a critical concern in these applications. It is crucial to prevent defects from arising in the first place, but it is equally necessary to detect and reduce the faults when they do arise. This is accomplished by using a combination of fault-tolerant architecture, redundancy, and error detection and correction systems in both hardware and software designs. Typically, measures like the probability of failure on demand (PFOD), the probability of dangerous failure per hour (PFH), or the safety integrity level are used to quantify fault occurrence in safety-critical systems (SIL). These metrics offer a numerical assessment of the level of safety that the system is capable of achieving and are applied to direct the design and testing procedures. This thesis aimed to figure out the reliability of application execution on a processor with and without an operating system and choose which one is more reliable for safety-critical applications. To this end, In order to examine the impact of radiations hitting the hardware while an application is being run, a system-on-chip architecture has been examined. Furthermore, two platforms have been produced to cover the aim of using a processor with the operating system and without. Then a fault injection platform has been developed with the goal of injecting faults in the main memory of RAM. The idea is to replicate radiation-related errors while an application is running. SEU (single even upset) has been chosen as the fault model for this experiment. Two programs from the MiBench benchmark suite and two other programs from a random source have been used as test input for the radiation impacts of an investigation. For the injection result, SDC, Halt, and Empty files have been chosen as the output classifications. It was possible to see that the error rate in some programs running on processors with operating systems is greater than the error rate in the same applications running on processors without operating systems by comparing the results given by the fault injection platforms. Therefore, it may be concluded from this conclusion that reliability could alter in the same program operating on a different system. So, in order to decrease the probability that an error would occur, we must choose an application carefully and supply it with a proper system.