Method for Distributed Redundant Execution of Program Modules
Background: Several schemes have been implemented in distributed computer systems which recover execution of program modules when hardware and software failures occur. The distributed recovery block scheme executes a primary and backup copy of a program on multiple processors. The processors communicate to determine which version failed such that the successful processor passes its result on to the next stage. Unfortunately, if one version crashes (fails to execute) the recovery time is prolonged. Additionally, the scheme is expensive to construct for processing more than two versions of a program. Another scheme, the autonomous decentralized software structure, executes more than two versions of the same program but is tolerant for hardware faults only. Technology: Scientists at the University of California have developed a new method for organizing and programming distributed computer systems which maintains the integrity of an application in the presence of multiple hardware and software failures in the processors. Parallel and asynchronous execution of multiple copies of the same or different versions of a program module is performed with processors which are connected by a network but do not interact directly during the execution. For each of the distributed program modules, a primary and backup version is executed concurrently by multiple processors. The results are checked and correct data are sent via transmission paths. Received data are checked for logical acceptability and sorted according to the data's program module of origin. Application: This scheme may be used in many situations including:- real time distributed computer control systems - non-real time commerical transaction processing systems
ADVANTAGES:
- Minimal delay of execution in the face of hardware or software failures - High tolerance for software and hardware faults
Patents:
US 5,551,047
Type of Offer: Licensing
« More Software Patents