The automatic detection of errors gives considerable protection against serious errors. With batch processing systems this may be sufficient, since there is ample time available for the manual diagnosis of the error and its eventual elimination. In contrast to this, real-time systems will rarely allow for manual intervention since the time constraints may be very tight. Thus mere error detection is insufficient in such a situation. As was stated in the previous chapter automatic error detection also has a negative influence on reliability. Precautions must therefore be taken with real-time systems to ensure that the system becomes operational again in the shortest possible time — often within a fraction of a second. The new operating level, which is aimed for after an error has occurred, in general will not be able to support all the tasks expected of a fully functional system. This is, however, not critical with the majority of systems, since the requirements for a system are divided into primary functions, which must be fulfilled under all circumstances, and secondary functions, which may be abandoned in the event of the occurrence of an error (see chapter 3). Clearly a system that automatically corrects errors must contain considerably more redundancy than would be required for mere detection. The amount of this additional redundancy determines which error types may be circumvented. While the correction of transient errors requires data and time redundancy as well as software for restarting, correction of permanent errors requires additional hardware components and software modules.
Unable to display preview. Download preview PDF.