Fault tolerant applications systems; A requirements perspective

Abstract

From the applications perspective, the requirements for fault tolerance are not intrinsic to the problem being solved. Fault tolerance requirements are at least two levels removed from the top level requirements; fault tolerance requirements are derived from the confluence of response time and availability requirements. Response time and availability requirements, in turn, are derived from several top-level requirements, most notably accuracy and performance requirements. This second-order requirements derivation results in a wide variety of responses in system architectures constructed to meet them.