Automatic recovery in a real-time, distributed, multiple microprocessor computer system

Abstract

Tnis tnesis presents an automatic recovery design that supports the fault-tolerant performance of a real-time, distributed, multiple microcomputer system. The recovery mechanism is stuctured to maintain real-time processing applications where a record of previous computations is not required and data loss is toleral)le during the period of recovery. The automatic recovery technique employed is based on system reinitialization in which the system is restored to it's original initialized state and then restarted. T^^e automatic recovery mechanism has been integrated with a hierarchical, distributed operating system which supports a mult ipro!?ramming" environment. A distinct address space for each system process, that is preserved by the hardware's explicit memory segmentation, in conjunction with the independent kernel and user domains of the operating system are used to facilitate dynamic relocation amon?: identical processor modules. The result is a flexible environment that supports the dynamic reconfiguration of processors and memory during the period of reinitialization.