Abstract: A data processing system is disclosed in which each software process monitors every other software process for failures. The detection of a failure can then be used to initiate any desired failure recovery strategy. Software failure detection takes place by having each software process assigned an imaginary resource by a resource locking mechanism. All other software processes request an assignment to the same imaginary resource. Whenever a software process fails, the assigned imaginary resource is released by the failed process and hence is available for assignment to the other software process. This availability of another process' resource is interpreted as a failure of that process, and initiates failure recovery. No actual resources are used, and a standard resource allocation mechanism, such as is used to assign shared read and exclusive write access to a file, is used to perform the locking functions.