Violation detection and recovery of distributed programs' safety properties
P.-N. Lee · 2003
On the basis of communication-closed layers of a distributed program, the author suggests the use of a communication-closed layer as the syndicate unit for dynamically detecting and recovering from violations of safety properties as defined by the static portions of the specifications. A fault-tolerant layer is composed of several communication-closed layers known as alternate layers, each of which would provide the same consensus function; thus, if a layer violates some safety property, it is known exactly where to perform a rollback and switch an alternate layer. An approach called the intermediate processes scheme is also presented. The intermediate buffer processes are utilized to provide both the asynchronous computation among application processes and error recovery for their dedicating application process. The intermediate regional testers were utilized to form a hierarchy of distributed software systems and to detect most of the consensus errors within regions.>