Automatic Design of Reliable Systems Based on the Multiple-choice Knapsack Problem

Fault-Tolerant System Design, Electronic Design Automation, Multiple-choice Knapsack Problem, Fault Tolerance Property Estimation, Verification, High-Level Synthesis

This paper evaluates the practical usage of the Multiple-choice Knapsack Problem (MCKP) solver to automatically select the proper fault mitigation method for each component to maximize the overall fault tolerance of the whole system. The usage of the MCKP is placed into the context with our fault tolerance automation toolkit, the goal of which is to completely automate the process of fault-tolerant system design on a very general level. To achieve our goal, we present our research on Field Programmable Gate Arrays (FPGAs) for which we have developed the specific components in order to support their fault-tolerant design automation. In our particular case study, the MCKP method on the partitioned system was able to find the solution with 18% less critical bits compared to our previous approach, while even lowering the circuit size. The results indicate that by splitting the system into smaller components and applying the MCKP method, considerably better results in terms of critical bits representation can be achieved.