Fundamentals of Dependability

Abstract

This chapter provides an overview of a number of fault-tolerant design schemes suitable for nanometer system-on-chip applications. Fault tolerance is the ability of a system to continue error-free operation in the presence of an unexpected fault. There are many different fault-tolerant design schemes. Choosing the scheme to use in a particular design depends on what types of faults need to be tolerated such as temporary or permanent, single or multiple point failures, and the constraints on area, performance, and power. As technology continues to scale and circuits become increasingly prone to failure, achieving sufficient fault tolerance will be a major design issue. It starts with an introduction to the basic concepts in fault-tolerant design and the metrics used to specify and evaluate the dependability of the design. Fault tolerance requires some form of redundancy in time, hardware, or information. An instance of time redundancy would be to perform the same operation twice and see whether the same result is obtained both times (if not, then a fault has occurred). This can detect temporary faults but not permanent faults, as a permanent fault would affect both computations. The advantage of time redundancy is that it requires little or no extra hardware. However, the drawback is that it impacts the system or circuit performance. Coding theory is also reviewed and some of the commonly used error detecting and correcting codes are described. This is followed by some examples of various types of fault-tolerant applications used in industry. © 2008 Elsevier Inc. All rights reserved.