EVALUATION OF HFT TECHNIQUE PERFORMANCE USING TASK-SCHEDULING ALGORITHMS

In earlier work we have introduced an efficient hardware fault-tolerant approach for reliable execution of tasks. The proposed approach called the Hardware Fault-Tolerant (HFT) approach. Also, we have introduced the concept of dynamic group maximum matching, which is used to group nodes of a graph into disjoint groups with different sizes dynamically. Furthermore, we have proposed the Dynamic Group Maximum Matching (DGMM) algorithm for finding the dynamic group maximum matching. In addition, we have proposed several hardware fault-tolerant scheduling algorithms, based on the HFT technique and the DGMM algorithm. In this work, we studied the effect of the HFT technique on system performance for four of the proposed scheduling algorithms. The studied algorithms are: Hardware Fault-Tolerant (FCFS + First Disagreement Graph First + First Started First), Hardware Fault-Tolerant (FCFS + First Fit First + First Disagreement Graph First + First Started First), Hardware Fault-Tolerant (FCFS + First Disagreement Graph First + Largest Group First), and Hardware Fault-Tolerant (FCFS + First Fit First + First Disagreement Graph First + Largest Group First) scheduling algorithms. Two performance metrics were evaluated. The first metric is system mean response time. The second metric is percentage of tasks of a certain type completed.