The problems associated with ensuring a correct and timely response are particularly acute in computing systems which must satisfy critical timing requirements. Faults must be recognised promptly and the overheads of state restoration and retries must be tightly controlled so that real-world time constraints are not violated. This paper describes how a form of rate-monotonic scheduling could be used to incorporate deadline mechanisms into transputer-based systems in order to provide timely fault-tolerance through conversations.
|Number of pages||8|
|Journal||Microprocessing and microprogramming|
|Publication status||Published - Sep 1993|