Stable paths for an efficient recovery in real-time distributed systems

This paper deals with the problem of communication in a distributed system, composed by nodes connected in a local area network. The system must continue its operation in the event of a node crash. In the system, agent functions cooperate and request services to server functions through message passing. The ability of server functions to provide continuous service, in presence of a node crash in a real-time and highly available environment, is obtained with the replication of the server functions in two different nodes (Master and Stand-by). A solution which allows a stable communication between agent and server functions is presented. This solution allows a role switching between the Master and the Stand-by without any loss of messages or need of re-entry of messages; this role switching is fully transparent to the agent functions.