Chapter 17 Action-level Fault Tolerance

Action-level (AL) fault tolerance means to accomplish every critical action (output action of a critical task as speciied) successfully in spite of component failures. Therefore, it is aimed for the highest degree of fault tolerance in real-time computer systems. Several basic techniques developed in recent years for realizing action-level fault tolerance in real-time LAN (local area network)-based systems are reviewed in this chapter. 17.1 Introduction Real-time computing is a relatively old branch (at least 35 years old) of computer engineering but has been advancing rather slowly until recently. In the author's opinion the main reason was the small market size for real-time computing. However , in recent years, the researcher population in the eld of real-time computing has shown rapid growth for various economic and technological reasons and as a result, the technology started advancing rapidly. Most large-scale real-time computing applications are of safety-critical type and thus require high reliability and fault tolerance properties from the computer systems used. The most desirable types of fault tolerance techniques for use in real-time computer systems (RTCSs) are those for realizing action-level (AL) fault tolerance , which means to accomplish every critical action (output action of a critical task as speciied) successfully in spite of component failures. Such techniques are thus aimed for the highest degree of fault tolerance in RTCSs. Several basic techniques developed in recent years for realizing action-level fault tolerance in real-time local area network (LAN)-based systems are reviewed in this chapter. The chapter starts with the author's personal perspectives of the evolution of the two elds, the RTCS eld and the fault-tolerant computer system (FTCS) eld, in Section 17.2. Then in Section 17.3, a classiication of various fault tolerance properties with respect to their values is discussed. In Section 17.4, two most impor-409