Fault-tolerant distributed algorithms for agreement and election
The first part characterizes completely the shared-memory requirements for achieving agreement in an asynchronous system of fail-stop processes that die undetectably. There is no agreement protocol that uses only read and write operations, even if at most one process dies. This result implies the impossibility of Byzantine agreement in asynchronous message-passing systems. The second part considers the election problem on asynchronous complete networks when the processors are reliable but some of the channels may be intermittently faulty. To be consistent with the standard model of distributed algorithms in which channel delay scan be arbitrary but finite, it is assumed that channel failures are undetectable. An algorithm is given that correctly solves the problem when the channels fail before or during the execution of the algorithm.
- Research Organization:
- Illinois Univ., Urbana, IL (USA)
- OSTI ID:
- 7222471
- Country of Publication:
- United States
- Language:
- English
Similar Records
Authenticated algorithms for Byzantine agreement
Efficient algorithm for Byzantine agreement without authentication