Cloud application platforms, with their support for elastic resources
and software development lifecycle, are attracting more and more users wishing
to host web-based applications that have variable demand, yet consistent
performance requirements. Effective platform and application management is
mandatory in order for applications running in the cloud to be cost-effective.
We see self-management as one of the way of achieving this goal. In this paper
we present our vision of an ontology-driven approach to self-management in
cloud application platforms. In the context of the MAPE-K reference model for
adaptation cycles, we describe main advantages of using ontologies and rules,
in order to represent self-reflective knowledge and define adaptation policies in
comparison to more traditional approaches.