Application Transparent Fault Management in Fault Tolerant Match

Fault detection and fault tolerance has become an increasingly important aspect of all computer system designs, from PC’s to high- end workstations and embedded critical systems. Since operating systems are common to all computers and it is at the operating system level where there is maximum system visibility and control, it is appropriate for the operating system to provide policies which detect, contain and tolerate faults. These policies form an operating system’s “fault management.” A mechanism to provide support for operating system fault management has been designed and implemented for a UNIX 43 BSD server running on the Mach 3.0 microkernel. The mechanism, called the sentry mechanism, consists of fault management control placed at all operating system entry and exit points. The suitability of the mechanism is determined through demonstration of its ability to support diverse, commonly accepted policies efficiently, where efficiency is measured in terms of implementation complexity and performance. Several sentry policies have been implemented including monitoring, assertions, checkpoint/checkpoint recovery and journaling journal replay. This paper presents the sentry mechanism, its implementation and the design and implementation of the mentioned policies.