Abstract:

In the research and application of multi-cross channel model, to maximize fault recovery of individual channel is the basis of the correctness to vote. There is some time redundancy in a task period. For a task processing in a given step, to summarize the time redundancy of pre-voting step, and assume fault-free on succedent step, then there will be a time redundancy on succedent step. The redundancy time of previous and succedent steps was counted, then a superior time window was used to do more deep recovery of fault. Based on the above ideas, a dynamic time series of multi-cross channel model was proposed, which was analyzed for deep recovery, and a backward recovery algorithm was given, which endowed more time to the fault unit, then the instantaneous fault could be eliminated to the utmost. Moreover, a monitoring logic was put forward to support the recovery algorithm. Theoretical analysis and experiments show that the backward recovery algorithm is effective to enhance the recovery rate and to reduce in the number of steps falling out. Compared with the statical recovery, the recovery rate increased by 47.49% and 72.35% respectively, and the number of out of step decreased by 58% and 85% respectively in the condition of 4 channel and 6 channel, which boosts the reliability of multi-cross channel model, especial in the condition of a large number of voting steps.