Automatically-Designed Fault-Tolerant Systems: Failed Partitions Recovery

Fault-Tolerant System Design, Electronic Design Automation, Redundancy Insertion, Redundancy Allocation, Multiple-choice Knapsack Problem, FPGA, VHDL, t50

This paper presents and describes our design automation toolkit for automatic synthesis of fault tolerant systems from unhardened systems. The toolkit is composed of various parts and tools and its aim is to design its internal algorithms in such way to be reusable among different HW description languages. In this paper, VHDL description is used to present the possibilities of the toolkit. The experimental part of the paper presents automatic synthesis of a benchmark system into a limited chip area. The optimization goal was to maximize the median time to failure (a.k.a. t50) parameter. The main part of the experimental activities comprises incorporation of a partial dynamic reconfiguration controller into the system design to recover the selected component of the system. Two systems utilizing recovery with the usage of the FPGA dynamic reconfiguration technique show promising results in terms of reliability. The recovered system, in which the controller is apart of the FPGA (e.g. in a different radiation-hardened chip), achieves by 70% better t50 parameter, compared to the system without recovery.