• analysis
• cloud applications
• management protocols
• modelling
• TOSCA

How to deploy and flexibly manage complex composite applications across heterogeneous cloud platforms is one of the main concerns in enterprise IT. Vendor-agnostic models to specify the structure and management of composite cloud applications, as well as techniques for the analysis of specified applications, would permit verifying the accuracy of the design of such applications. Furthermore, the availability of techniques for reusing composite cloud applications would free developers from the need of designing and developing multiple times recurring application components.

The objective of this thesis is to propose a suitable representation for composite cloud applications, and to develop analysis and reuse techniques based upon, but not limited to, such representation. By relying on TOSCA (Topology and Orchestration Specification for Cloud Applications) as the reference language for syntactically describing composite cloud applications, we propose a simple modelling that permits specifying the behaviour of the management operations of an application's components. We show how their management behaviour can be modelled by management protocols}, specified as finite state machines whose states and transitions are associated with conditions constraining the consistency of states and the executability of operations. We illustrate how to derive the management behaviour of a composite cloud application by composing the protocols of its components, and how this permits automating various analyses concerning the management of a cloud application. Last, but not least, we show how management protocols can be easily extended to take also into account the potential occurrence of faults while managing composite cloud applications.

Additionally, to enact the reuse of existing solutions, we illustrate how to syntactically matchmake (fragments of) TOSCA-based applications with respect to desired components, and how to adapt matching applications to concretely implement such components. We define two different notions of simulation between management protocols, and we exploit such notions to extend the proposed matchmaking and adaptation approaches by including the behaviour information in management protocols.