Dependability-Explicit Computing: Applications in e-Science and Virtual Organisations

Abstract

Providing a predictable level of dependability is a challenge for applications which choreograph services from many different providers. Applications commonly fail because a component service fails, yet the designers of applications have, at best, limited information about component service dependability. This limits their ability to make informed decisions about when it is cost-effective to use a service or to employ potentially expensive fault containment or tolerance techniques such as redundancy. We consider ways to improve support for the publication and exploitation of dependability metadata for services by developing publication methods and ontologies to support shared metadata definitions. Two diverse examples of metadata are considered: service availability information and descriptions of service failure modes. The availability work is particularly relevant to bioinformatics, while work on failure modes is explored in the context of virtual organisations with long-term interactions.