Abstract

In a system based on the imprecise-computation technique, each time-critical task is designed in such a way that it can produce a usable, approximate result in time whenever a failure or overload prevents it from producing the desired, precise result. This section describes ways to use this technique together with traditional fault-tolerance methods to reduce the costs of providing fault tolerance andenhancedavailability. Specifically, an imprecise mechanism for the generation and use of approximate results can be integrated in a natural way with traditional checkpointing and replication mechanisms. Algorithms and process structures for this integration and rules for determining when approximate results can be used in place of the desired results are discussed.

Citations