| 摘要 |
<p>A fault tolerant computer system which operates to automatically provide fault tolerant operation in a manner which reduces the frequency of checkpointing, while requiring neither simultaneity nor atomicity with respect to information transmitted to backups. These advantage are accomplished by providing operation such that message transmissions from primary tasks running on a data processor to external backups may be delayed until an action is performed which will persist after failure of the data processor. Although such operation may result in the backups hot having received all of the messages transmitted to their respective primary tasks when failure occurs, recovery is nevertheless obtained using the backups, since transmission to backups is controlled so that all backups agree on the state of the data processor at the time of failure, which need not be the actual state of the data processor at the time of failure.</p> |
