Wave III: Runtime Models

Runtime models can be considered as an extension of model‐driven engineering into the runtime. They provide an answer to the need to manage the complexity of concrete designs of self‐adaptive systems. This chapter provides a definition of a runtime model in self‐adaptive systems and elaborates on the motivations for runtime models. It discusses four key dimensions of runtime models: structural versus behavioral, declarative versus procedural, functional versus qualitative, and formal versus informal. The chapter describes three principal strategies that can be applied to using runtime models in self‐adaptive systems: MAPE components share K models, MAPE components exchange K models, and MAPE models share K models. In the first strategy, MAPE components share knowledge models. In the second strategy, MAPE components exchange runtime models to realize self‐adaptation. In the third strategy, the MAPE functions are realized as runtime models themselves, similar to the knowledge models they share.