Naslov
Experimental Evaluation of Dependability of Standby Fault-Tolerant Microcomputer System by Fault Injection

Autori
Hocenski, Željko ; Veizer, Berislav

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni

Izvornik
Proceedings KoREMA '96, 41'st Annual Conference, Vol. 4 / Perić, Nedjeljko - : Hrvatsko društvo za komunikacije, računarstvo, elektroniku, mjerenja I automatiku (KoREMA), 1996, 67-70

Skup
41'st Annual Conference KoREMA

Mjesto i datum
Opatija, Hrvatska, 18.09.1996. - 20.09.1996

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
fault tolerance; standby; fault injection; fault detection coverage; fault recovery coverage; microcomputer system

Sažetak
The fault-tolerant system based on cold spare standby is made by microprocessor modular systems. The dependability of this system is evaluated by using the fault injection method. A microcomputer system consists of standard nonredundant modules and specific additional modules, such as bus interfaces and a supervision unit. The supervision unit performs the validation of the active unit by the watch-dog timer, it generates spare actualization signals and controls the bus interface transfere directions. This unit can stop the active unit in case of the detected fault, and transfers the control to a spare unit. The fault injection is performed by injection into the bus lines of the system under test for certain adjustable time duration. The fault injector equipment consists of one personal computer and additional extended hardware. It performs the fault injection and records the system reaction, repeating that in order to get some statistical data. The additional hardware for personal computer and software are the results of our own research. The used fault injection method performs the flexible location for injection, variable impulse duration, and the choice of the injected signal logical level. The results of testing are given as the number of detected faults of the active unit and the number of failures of the whole system for two logic levels of faults injected to bus lines as a function of fault duration. The activity on bus lines is registered and the probability of fault occurrence is calculated. The fault detection coverage and fault recovery coverage are evaluated by using the probability of occurrence and the registered number of faults. Two types of software are tested: one nonredundant and one redundant version. The recovery coverage for the system with redundant software is much better than for the system with nonredundant software, but many of faults are masked, so that fault detection coverage is not well known for the system with redundant software.

Izvorni jezik
Engleski

Znanstvena područja
Elektrotehnika

POVEZANOST RADA