Tecniche di fault tolerance in sistemi concorrenti

In this paper the problem resulting from the application of fault tolerant techniques to concurrent programming field are presented. Firstly, the fundamental techniques adopted in sequential programming are briefly recalled. Particularly, the "recovery block" concept is introduoed and the "recovery cache" mechanism, especially suitable for its implementation, is presented. Secondly, the problem of stating recovery actions in the case of concurrent systems is disoussed by separatley analysing the two cases of process interaction: competition and cooperation. As far as competition is concerned, the recovery actions are discussed in the hypothesis that error conditions are detected during or after the resource acquisition, use or release. As far as cooperation is concerned, some recovery schemes are presented, able to avoid the danger that all cooperating processes must return to the initial recovery point when an error condition is detected.