Automatic reconnection of linked software processes in fault-tolerant computer systems

A join manager 110 controls the post-failure reconnection of software processes A and B in a fault-tolerant computer system in which each software process has a running backup process 122,132. The link before failure is denoted 160. Because there is a join manager, the software processes do not need to maintain knowledge of the redundancy strategy of their partner process. The join manager also enables reconnection of linked processes running in different parts of a heterogeneous distributed system. No polling mechanism is required in the partner processes to achieve the reconnection. This saves processor and communication resources by avoiding repeated attempts to poll a failed partner to check if it has recovered. Instead of running each software process on a separate CPU, more than one of the software processes may be run on the same CPU.