A Fault-Tolerant Scheduling Method Based on Predictable Deadline Miss Ratio in High Utilization

Modern real-time systems, such as navigation and communication, emphasize high reliability and validity, which requires 100% error detection and fault recovery. Meanwhile, those systems also face the requirement of sophisticated real-time digital signal processing and information intercommunication, where real-time processing is in high utilization, resulting in a lack of timing resource for error detection and fault recovery. This requires that systems have predictable performance. Traditional time redundancy methods cannot deal with high utilization, since they need to guarantee fault recovery while no deadline is missed, which severely restricts their use in high utilization systems. Researches show that traditional time redundancy fault-tolerance in high utilization real-time systems may result in a disaster. A new fault-tolerant method is presented to deal with the problem. The main contribution of this paper is that a fault-tolerant method with predictable DMR(deadline miss ratio) in high utilization is proposed. The number of deadline missing is not greater than the number of error occurrences, which eliminates the domino effect of continuous deadline missing. A further improved approach is presented, and DMR can be effectively reduced by adopting time redundancy techniques based on off-line checkpointing analysis. The simulation experiments show that the proposed method leads to lower predictable DMR compared with the well-known algorithms so far.