Language support for service-level agreements for application-service provision

My thesis is that practical language support can be provided for Service-Level Agreements (SLAs) for Application-Service Provision (ASP), which is better than that provided by pre-existing languages in that: it provides greater assistance in expressing conditions that mitigate the risks inherent in ASP; and disputes related to agreements expressed in this manner may be more easily resolved in so as to respect the original intent of the parties. I support this thesis by establishing requirements for SLAs for ASP based on an account of a typical ASP infrastructure and business model. These identify the particular risks inherent in ASP, permit comparisons between ASP SLA languages, and guide the development of an abstract, extensible, domain-specific language, SLAng. SLAng is defined using a meta-modelling approach that allows a high degree of precision in the specification of its semantics, traceability from SLA to language specification, and the testing of the language and SLAs to ensure they capture the original intent of the parties. SLAng supports the expression of mutually-monitorable SLAs, for which the determination of compliance depends only on events visible to both client and provider of the service. I demonstrate that such SLAs are the most monitorable possible in a typical ASP scenario, given current monitoring technology, and describe an approximately-monitorable constraint on the accuracy of evidence used to administer such SLAs. SLAng is shown to be of practical use in a case study, evaluated against the original requirements, and compared with pre-existing languages. The evaluation of SLAng is enhanced using metrics developed to assist in assessing the contribution of a domain-specific language specification to encoding the meaning of statements in that language.