Biased Voting: Proposed Novel Error Detection and Recovery Mechanism

SUMMARY & CONCLUSIONSFault tolerance and redundancy are often used to increase system lifetime. Several identical copies of the same component are used in the fault-tolerant architecture. One of the main disadvantages of redundancy is obviously the extra cost. Especially in developing countries, this extra cost may be prohibitive. One option is to use several functionally identical components but with different failure rates. Lower cost components typically have a higher failure rate. However, the cost of the fault-tolerant architecture will be lowered.In this paper, such a system is studied. Four functionally identical components are used; three of them have a low cost and a relatively high failure rate while the fourth component is expensive and has a relatively low failure rate. The proposed Error Detection and Recovery Mechanism (EDRM) uses a biased voting scheme between operational and faulty devices. The fault model used in the design is single and double module failures. Using Continuous Time Markov Models (CTMCs), it is shown that the proposed system with biased voting has a reliability that is much higher than the simplex system and is only a little lower than a conventional sift-out system; however, the cost of the proposed system is much lower than that of the sift-out system. Furthermore, it is proven that the proposed technique can be used to increase system reliability without having to replace all redundant modules by more reliable ones. Future work might include the addition of rewards to the failure states.Finally, the EDRM is implemented as a ROM-based Finite State Machine hosted by an Intel CPLD chip MAX II EPM2210F324.