Recovery blocks for communicating systems

In many practical applications of real-time computing (avionics, switching systems) a message-passing inter-processes communication approach is adopted for both modularity and reliability aims. In the present paper, the problem of adding fault-tolerance in a message passing multiprocesses environment is examined. Recovery blocks implementation schemes for both asynchronous and synchronous communications are proposed, with the aim of avoiding domino-effects and exploiting the message oriented system structure. When a sender process produces a message, an acceptance test is performed on the message by system procedures, which in sequence: i) transfer the message on the receiving process working memory, ii) save present process status, or in case of error, restore some previous process status, and iii) discard no longer needed status informations.